# varg how to > AI Video Skills for Claude Code, Codex and Grok. How to make any video with AI agents: step-by-step tutorials, and 582 AI video prompts, effects, styles, guides, workflows and model skills from varg, HeyGen HyperFrames, Remotion, fal, ElevenLabs, Higgsfield, Krea, Runway, inference.sh and independent creators, each with exact install commands for Claude Code, Codex, Cursor, OpenClaw and Gemini CLI. The generation itself runs on varg (one API key for Seedance 2.5, Kling 4, Veo 3.1, ElevenLabs and 200+ models) or on the vendor's own service. How an agent uses this site: 1. Search in one call: `curl -s "https://varg.ai/how-to/api/search?q=ugc+ad"` returns the best how-to page and the top skills with install commands, as markdown. JSON: `https://varg.ai/how-to/api/search.json?q=…`. Russian queries work. 2. Open the how-to page it names. It lists the best skill, alternatives, the visual options to show the person before spending anything, and the steps. 3. Show the person 2–4 options (looks, formats, voices) with the example links, then install the chosen skill and run it. 4. Everything machine-readable: https://varg.ai/how-to/catalog.json. Each skill installs with the command on its page (its own repo). `npx skills add https://varg.ai/how-to` installs only varg's own skills (varg-ai and friends). ## How to: step-by-step tutorials - [How to make a video with AI](https://varg.ai/how-to/make-a-video.md): Start from the video you need: product videos, ads and UGC, Shorts and Reels, cinematic films, short dramas, animation and music videos. Each rubric gathers the workflows, effects, prompts and models that make it. - [How to edit a video with AI and FFmpeg](https://varg.ai/how-to/edit-a-video.md): Start from footage you already have. The editing agents and pipelines that cut a whole video; every way to cut and join two shots, CapCut-style and in FFmpeg, Remotion and HyperFrames; the words on screen (headlines, captions, quotes, end cards); how to show a product (UI, cursors, devices, code, agent flows); and the visuals around it (numbers, logos, grids, backdrops). Each take says what it looks like, when to use it, the common mistake and how to build it. - [How to make AI images and photos for your video](https://varg.ai/how-to/make-ai-images-and-photos.md): Images for your video and around it: YouTube thumbnails, consistent characters and locations, style references, art styles, logos and brand kits. - [How to make AI music, voiceover and audio for video](https://varg.ai/how-to/make-ai-music-voiceover-and-audio.md): Music, voice and sound for your video: AI songs and lyrics, text to speech, voice cloning, dubbing and sound effects. - [How to make a UGC ad video with AI](https://varg.ai/how-to/make-a-ugc-ad-video.md): A creator-style ad: a person talks to camera about your product, with b-roll, captions and an end card. - [How to make video ad variants for testing with AI](https://varg.ai/how-to/make-video-ad-variants-for-testing.md): Ten or more short video ads from one brief, each with a different hook or format, for Meta/TikTok creative testing. - [How to turn a product photo into a video with AI](https://varg.ai/how-to/turn-a-product-photo-into-a-video.md): A product shot turned into a short promo: orbit, push-in, exploded view or bullet time, with music. - [How to cut a podcast into short video clips with AI](https://varg.ai/how-to/cut-a-podcast-into-short-video-clips.md): Find the best moments in a long video, cut them, reframe to 9:16 and burn captions. - [How to add animated captions to a video with AI](https://varg.ai/how-to/add-animated-captions-to-a-video.md): TikTok-style subtitles burned into an existing video: word highlight, pop words, or designed editorial captions. - [How to make a talking avatar video with AI](https://varg.ai/how-to/make-a-talking-avatar-video.md): A person (from a headshot, or generated) speaks your script to camera with lip-sync, in any language. - [How to translate and dub a video with AI](https://varg.ai/how-to/translate-and-dub-a-video.md): Your video in another language with your own voice, optionally lip-synced and subtitled. - [How to make a faceless YouTube video with AI](https://varg.ai/how-to/make-a-faceless-youtube-video.md): A narrated short or long video with no presenter: script, voiceover, visuals, captions, music. - [How to make a text message story video with AI](https://varg.ai/how-to/make-a-text-message-story-video.md): A viral iMessage-style chat drama: bubbles appear with typing, sounds and a cliffhanger. - [How to make a Notes app video ad with AI](https://varg.ai/how-to/make-a-notes-app-video-ad.md): A note types itself in a phone notes app, reads like a confession, and turns into an ad. - [How to make a SaaS launch video with AI](https://varg.ai/how-to/make-a-saas-launch-video.md): A polished launch film for an app or website: UI cards, device mockups, kinetic type, music. - [How to turn release notes into a changelog video with AI](https://varg.ai/how-to/turn-release-notes-into-a-changelog-video.md): Release notes, a PR or a GitHub repo turned into a short video for X or LinkedIn. - [How to make an explainer video with AI](https://varg.ai/how-to/make-an-explainer-video.md): A narrated explainer about a product, a topic or a document, with motion graphics or AI visuals. - [How to make an AI music video](https://varg.ai/how-to/make-an-ai-music-video.md): A music video for your song: concept, a consistent lead character, shots cut on the beat, lip-sync on vocals. - [How to make a lyric video with AI](https://varg.ai/how-to/make-a-lyric-video.md): Lyrics synced to the song: the sung line big and highlighted, karaoke-style, over art or a shader. - [How to turn photos into a slideshow video with AI](https://varg.ai/how-to/turn-photos-into-a-slideshow-video.md): Your photos cut to music, with motion, transitions on the beat and titles. - [How to make an AI short film](https://varg.ai/how-to/make-an-ai-short-film.md): A multi-shot story with consistent characters: script, storyboard, previz, shots, edit. - [How to keep a character consistent in AI video](https://varg.ai/how-to/keep-a-character-consistent-in-ai-video.md): One mascot, trainer or actor that looks identical across many images and clips. - [How to make an AI influencer video](https://varg.ai/how-to/make-an-ai-influencer-video.md): A recurring AI character with one look and one format, posting as a creator or a brand face. - [How to make exercise demo video loops with AI](https://varg.ai/how-to/make-exercise-demo-video-loops.md): Looping exercise demos for a fitness app: one trainer, many exercises, clean background. - [How to make a logo animation with AI](https://varg.ai/how-to/make-a-logo-animation.md): Your logo animated into a short intro or outro: 3D chrome, liquid, shader, glitch or clean motion. - [How to make an animated chart video with AI](https://varg.ai/how-to/make-an-animated-chart-video.md): Numbers from a CSV or a report as an animated chart: bar race, line, hero number. - [How to make YouTube video thumbnails with AI](https://varg.ai/how-to/make-youtube-video-thumbnails.md): High-CTR thumbnails and vertical covers, in a few options to choose from. - [How to make AI product photos and listing images](https://varg.ai/how-to/make-ai-product-photos.md): Marketplace listing images, lifestyle shots and packshots from one product photo. - [How to edit your own video footage with AI](https://varg.ai/how-to/edit-your-own-video-footage.md): Cut silences and flubs, reframe 16:9 to 9:16, add captions and b-roll to footage you shot. - [How to remake a reference video with AI](https://varg.ai/how-to/remake-a-reference-video.md): Take a reel or ad you like, break it down shot by shot, and remake the format for your brand. - [How to add voiceover and music to a video with AI](https://varg.ai/how-to/add-voiceover-and-music-to-a-video.md): A professional voiceover and a fitting music bed added to your video or screen recording. - [How to animate a photo into a video with AI](https://varg.ai/how-to/animate-a-photo-into-a-video.md): A still photo brought to life: a subtle smile, blink, camera move or full scene motion. - [How to make trending AI video formats](https://varg.ai/how-to/make-trending-ai-video-formats.md): What's going viral now — formats you can make for your brand today, with the skill for each. - [How to swap your character into a video with AI](https://varg.ai/how-to/swap-your-character-into-a-video.md): Your character (or person) performs the exact motion of a trending video: dance, gesture, walk, ad performance — motion transfer. - [How to make AI b-roll for your video](https://varg.ai/how-to/make-ai-b-roll-for-your-video.md): Cutaway footage for every mention in a voiceover or interview: generated clips, collage or stock-style shots, cut in on time. - [How to make TikTok photo carousels with AI](https://varg.ai/how-to/make-tiktok-photo-carousels.md): Swipeable photo slideshows with hook text — the format app founders use for organic growth. - [How to make an app demo video with AI](https://varg.ai/how-to/make-an-app-demo-video.md): A polished demo of your app: device mockups, real screens, captions, music — for the App Store, ads or a landing page. - [How to make a property listing video with AI](https://varg.ai/how-to/make-a-property-listing-video.md): A listing reel from property photos: gentle camera motion through rooms, captions with price and features, music. - [How to change something inside a video with AI](https://varg.ai/how-to/change-something-inside-a-video.md): Swap a product, restyle a scene, change lighting or the look of an existing clip while keeping the motion. - [How to make personalized videos at scale with AI](https://varg.ai/how-to/make-personalized-videos-at-scale.md): One template, many versions: each lead's name, company and website in the video, optionally in a cloned voice. ## Best tools, compared - [Best AI video editing tools for agents in 2026: 35 compared](https://varg.ai/how-to/best-ai-video-editing-tools.md): From your footage to the right tool: 35 AI video editing tools for agents compared by price, access and where they run. Pick: FFmpeg. - [Best AI video clipping tools in 2026: 3 compared](https://varg.ai/how-to/best-ai-video-clipping-tools.md): From your footage to the right tool: 3 AI long video to shorts tools for agents compared by price, access and where they run. Pick: OpusClip. - [Best AI video caption tools in 2026: 4 compared](https://varg.ai/how-to/best-ai-video-caption-tools.md): From your footage to the right tool: 4 AI captions & subtitles tools for agents compared by price, access and where they run. Pick: FFmpeg + ASS. - [Best AI transcript video editors in 2026: 4 compared](https://varg.ai/how-to/best-ai-transcript-video-editors.md): From your footage to the right tool: 4 AI talking-head cleanup tools for agents compared by price, access and where they run. Pick: ElevenLabs Scribe. - [Best AI video editing models in 2026: 4 compared](https://varg.ai/how-to/best-ai-video-editing-models.md): From your footage to the right tool: 4 AI ai edits: restyle, swap, relight tools for agents compared by price, access and where they run. Pick: varg video edit - [Best AI video dubbing tools in 2026: 3 compared](https://varg.ai/how-to/best-ai-video-dubbing-tools.md): From your footage to the right tool: 3 AI dubbing & lip sync tools for agents compared by price, access and where they run. Pick: HeyGen Video Translate. - [Best AI video upscaling tools in 2026: 3 compared](https://varg.ai/how-to/best-ai-video-upscaling-tools.md): From your footage to the right tool: 3 AI upscale, slow-mo, cut-outs tools for agents compared by price, access and where they run. Pick: Topaz Video API. - [Best video editing APIs for AI agents in 2026: 4 compared](https://varg.ai/how-to/best-video-editing-apis-for-ai-agents.md): From your footage to the right tool: 4 AI assemble from a timeline tools for agents compared by price, access and where they run. Pick: HyperFrames. ## With FFmpeg: copyable commands - [How to burn captions into a video with FFmpeg](https://varg.ai/how-to/burn-captions-with-ffmpeg.md): Turn a video and an SRT file into an MP4 with the captions burned into the picture, in one FFmpeg command. Free and local. - [How to crop a video to 9:16 vertical with FFmpeg](https://varg.ai/how-to/crop-a-video-to-vertical-with-ffmpeg.md): Turn a 16:9 video into a 1080×1920 vertical 9:16 MP4 for TikTok, Reels and Shorts, with one FFmpeg command. Free and local. - [How to trim and join video clips with FFmpeg](https://varg.ai/how-to/trim-and-join-video-clips-with-ffmpeg.md): Turn long footage into trimmed clips and join them into one MP4 with FFmpeg: the exact commands for frame-accurate cuts and lossless joins. Free and local. - [How to add music to a video with FFmpeg](https://varg.ai/how-to/add-music-to-a-video-with-ffmpeg.md): Turn a video and a music track into one MP4 with the music under the voice, ducked while people talk, with one FFmpeg command. Free and local. ## Skills by type and by job - [Prompts](https://varg.ai/how-to/prompts.md) (29): AI video prompts with examples: Seedance, Kling, Veo - [Effects](https://varg.ai/how-to/effects.md) (231): AI video effects: titles, captions, transitions - [Styles](https://varg.ai/how-to/styles.md) (14): AI video styles: film looks, art styles, LUTs - [Guides](https://varg.ai/how-to/guides.md) (112): AI video tutorials and best practices: Remotion, FFmpeg - [Workflows](https://varg.ai/how-to/workflows.md) (78): AI video generator workflows: ads, UGC, product, music - [Models & tools](https://varg.ai/how-to/models.md) (118): AI video models: how to use Seedance, Veo, Kling, FLUX - [Viral prompts with results](https://varg.ai/how-to/prompts/viral.md) (35): full prompts from X for GPT Image 2.5, Claude Opus 5.5, Seedance 2.5, Nano Banana 2.1 and more, with the result each made - [Trending styles](https://varg.ai/how-to/styles/trending.md) (32): the looks people ask for now, each with a prompt clause and the best models - [Make a video](https://varg.ai/how-to/make-a-video.md): [Product videos](https://varg.ai/how-to/browse/skills/product.md), [Ads & UGC](https://varg.ai/how-to/browse/skills/ads.md), [Shorts & Reels](https://varg.ai/how-to/browse/skills/shorts.md), [Cinematic & short films](https://varg.ai/how-to/browse/skills/cinematic.md), [Stories & short dramas](https://varg.ai/how-to/browse/skills/story.md), [Animation & motion graphics](https://varg.ai/how-to/browse/skills/animation.md), [Music videos](https://varg.ai/how-to/browse/skills/music.md), [Launch videos](https://varg.ai/how-to/browse/skills/launch.md), [Explainers](https://varg.ai/how-to/browse/skills/explainers.md), [Talking heads & avatars](https://varg.ai/how-to/browse/skills/talking-head.md), [Text & titles](https://varg.ai/how-to/browse/skills/kinetic-type.md), [Logo animations & intros](https://varg.ai/how-to/browse/skills/brand-motion.md), [3D scenes](https://varg.ai/how-to/browse/skills/3d-scenes.md), [Phone & app screens](https://varg.ai/how-to/browse/skills/native-ui.md), [Greetings & celebrations](https://varg.ai/how-to/browse/skills/greetings.md) - [Edit a video](https://varg.ai/how-to/edit-a-video.md): [AI editing agents](https://varg.ai/how-to/browse/skills/edit-agents.md), [Editing services](https://varg.ai/how-to/browse/skills/edit-services.md), [Editing & clips](https://varg.ai/how-to/browse/skills/editing.md), [Dubbing & lip sync](https://varg.ai/how-to/browse/skills/dubbing.md), [Cuts & pacing](https://varg.ai/how-to/browse/skills/cuts.md), [Transitions](https://varg.ai/how-to/browse/skills/transitions.md), [Headlines & callouts](https://varg.ai/how-to/browse/skills/titles.md), [Captions & subtitles](https://varg.ai/how-to/browse/skills/captions.md), [Quotes & reviews](https://varg.ai/how-to/browse/skills/quotes.md), [End cards & CTAs](https://varg.ai/how-to/browse/skills/end-cards.md), [UI walkthroughs](https://varg.ai/how-to/browse/skills/ui.md), [Cursors & clicks](https://varg.ai/how-to/browse/skills/cursors.md), [Device mockups](https://varg.ai/how-to/browse/skills/devices.md), [Code & terminal](https://varg.ai/how-to/browse/skills/code.md), [AI agent flows](https://varg.ai/how-to/browse/skills/agent-flows.md), [Numbers & charts](https://varg.ai/how-to/browse/skills/numbers.md), [Customer logos](https://varg.ai/how-to/browse/skills/logo-walls.md), [Grids & carousels](https://varg.ai/how-to/browse/skills/galleries.md), [Backdrops](https://varg.ai/how-to/browse/skills/backdrops.md) - [Make images](https://varg.ai/how-to/make-ai-images-and-photos.md): [Images & thumbnails](https://varg.ai/how-to/browse/skills/image.md), [Characters](https://varg.ai/how-to/browse/skills/characters.md), [Scenes & locations](https://varg.ai/how-to/browse/skills/locations.md), [Style references](https://varg.ai/how-to/browse/skills/style-refs.md), [Art styles](https://varg.ai/how-to/browse/skills/art-direction.md), [Logos & brand kits](https://varg.ai/how-to/browse/skills/logos.md) - [Make music & audio](https://varg.ai/how-to/make-ai-music-voiceover-and-audio.md): [AI music](https://varg.ai/how-to/browse/skills/ai-music.md), [Lyrics & songwriting](https://varg.ai/how-to/browse/skills/lyrics.md), [Voice & voiceover](https://varg.ai/how-to/browse/skills/audio.md), [Sound effects](https://varg.ai/how-to/browse/skills/sfx.md) - [Which service to edit a video with](https://varg.ai/how-to/browse/skills/edit-services.md): 10 edit jobs (cuts, transcript edits, clips, captions, AI video-to-video, dubbing, upscaling), the pick for each, alternatives and gotchas ## Video types made with code (recipes) Beat-by-beat recipes with the components that build each beat (1146 components: HyperFrames, remocn/Remotion, varg templates; each installs with one command — index: https://varg.ai/how-to/components.md). - [Product launch film](https://varg.ai/how-to/video/product-launch.md): The launch video for a product or a startup: hook, problem, features on real screens, proof, call to action. - [Feature announcement](https://varg.ai/how-to/video/feature-announcement.md): One new feature, shown working: the name, the click that matters, the benefit, where to try it. - [App promo & App Store preview](https://varg.ai/how-to/video/app-promo.md): An app sold on its own screens: devices, flows, ratings and the download badge. - [UGC ad with overlays](https://varg.ai/how-to/video/ugc-overlay-ad.md): Creator footage made to perform: a hook line, word captions, callouts on the product, an offer and a CTA. - [Ad variants for testing](https://varg.ai/how-to/video/ad-variants.md): One ad rendered as dozens: every hook, size and offer from a spreadsheet, same structure. - [Sale & offer promo](https://varg.ai/how-to/video/offer-promo.md): A price that drops, a code to copy and a clock that runs out. - [Testimonial & review ad](https://varg.ai/how-to/video/testimonial-ad.md): A real review read over the product: the quote, the stars, the numbers behind it. - [Us vs them & before / after](https://varg.ai/how-to/video/comparison-ad.md): Your product next to the usual way: a split frame, the numbers that differ, the switch. - [Notes-app ad](https://varg.ai/how-to/video/notes-ad.md): A message typed live in Apple Notes — the native look that reads as a friend, not an ad. - [Captioned talking clip](https://varg.ai/how-to/video/captioned-clip.md): Someone talking to camera, made watchable on mute: word captions, name tag, emphasis and punch-ins. - [Faceless narrated video](https://varg.ai/how-to/video/faceless-explainer.md): A voiceover carried by type, images and captions — no presenter, all code and stock. - [Top-N listicle](https://varg.ai/how-to/video/listicle.md): A countdown of items, each with an image, a number and one line, ending on the winner. - [Post-story narration](https://varg.ai/how-to/video/story-narration.md): A Reddit or forum post read aloud over looping gameplay, the post card first, captions throughout. - [Text-message story](https://varg.ai/how-to/video/chat-story.md): A story told as a message thread: bubbles arrive, typing dots tease, the last text lands. - [Post to video](https://varg.ai/how-to/video/post-to-video.md): A tweet, a thread or a comment turned into a clip: the card, the highlight, the reaction. - [Quiz, poll & would-you-rather](https://varg.ai/how-to/video/quiz.md): A question, options, a ticking clock and the reveal — interactive in a feed that is not. - [Meme format](https://varg.ai/how-to/video/meme.md): A known meme with your product as the punchline: setup, beat, payoff. - [Channel intro & end screen](https://varg.ai/how-to/video/channel-intro.md): The few seconds every video shares: the opener, the name tag, the subscribe nudge and the end screen. - [Changelog & release notes](https://varg.ai/how-to/video/changelog.md): A release as a video: version, what changed, the one change worth seeing, how to upgrade. - [Pull request to video](https://varg.ai/how-to/video/pr-video.md): A merged PR explained: the title, the diff that matters, before and after, the people who shipped it. - [Code walkthrough](https://varg.ai/how-to/video/code-walkthrough.md): Code read on screen: files open, lines light up, the camera follows the logic, the result runs. - [CLI & terminal demo](https://varg.ai/how-to/video/terminal-demo.md): A command line tool in action: the command types, the output streams, the result lands. - [Website to video](https://varg.ai/how-to/video/website-to-video.md): A live site turned into a promo: the hero in a browser frame, scroll-throughs, features, the CTA. - [App walkthrough & onboarding](https://varg.ai/how-to/video/app-walkthrough.md): How to do one thing in the app, step by step, with taps, zooms and callouts. - [AI chat & agent demo](https://varg.ai/how-to/video/ai-demo.md): An AI product shown the way people meet it: a prompt typed, the answer streaming, the result. - [API & docs explainer](https://varg.ai/how-to/video/docs-explainer.md): A concept from the docs drawn as a system: parts, flow, the code that calls it. - [Data story](https://varg.ai/how-to/video/data-story.md): One question answered with a chart that builds, annotations that point, a number that sticks. - [Bar chart race](https://varg.ai/how-to/video/chart-race.md): Rankings over time: bars overtake each other while a year ticks. - [Map & route animation](https://varg.ai/how-to/video/map-story.md): Places on a map: a camera flies in, a route draws, numbers land on the regions. - [News explainer](https://varg.ai/how-to/video/news-explainer.md): A story in 60 seconds: the headline, the context on a map, the numbers, what it means. - [Year in review / Wrapped](https://varg.ai/how-to/video/wrapped.md): A person’s own numbers as a story: top stats, top lists, a persona, a card to share. - [Scores & match recap](https://varg.ai/how-to/video/sports-recap.md): The result first, then the moments and the table: built from a match feed. - [Daily market or weather update](https://varg.ai/how-to/video/daily-update.md): The same video every morning with new numbers: rendered from an API, no editing. - [Math animation](https://varg.ai/how-to/video/math-animation.md): 3Blue1Brown-style: shapes construct the idea, equations transform step by step. - [Diagram explainer](https://varg.ai/how-to/video/diagram-explainer.md): How a system works: parts appear, arrows flow, the camera zooms where it matters. - [Whiteboard & hand-drawn](https://varg.ai/how-to/video/whiteboard.md): Ideas drawn as they are said: ink lines, scribbled arrows, handwritten titles. - [Step-by-step tutorial](https://varg.ai/how-to/video/tutorial.md): A how-to with numbered steps, the screen, the click, the tip, the finished result. - [Flashcards & language drills](https://varg.ai/how-to/video/flashcards.md): A word, its picture, its meaning, a pause to answer — a deck rendered from a word list. - [History timeline](https://varg.ai/how-to/video/history-timeline.md): Events along a line: dates tick, places light up, archive images drift. - [Science what-if explainer](https://varg.ai/how-to/video/science-explainer.md): A question at an impossible scale: 3D models, particles and numbers that keep growing. - [Lyric video](https://varg.ai/how-to/video/lyric-video.md): The song’s words on screen, timed to the vocal: type that moves on the beat over a look. - [Karaoke video](https://varg.ai/how-to/video/karaoke.md): Lines that fill word by word as they are sung, with a count-in and a backing picture. - [Audio visualizer](https://varg.ai/how-to/video/visualizer.md): The track made visible: spectrum, waveform or a shader that breathes with the bass. - [Podcast clip & audiogram](https://varg.ai/how-to/video/audiogram.md): A podcast moment for the feed: the waveform, the captions, who is speaking, where to listen. - [Beat-synced slideshow](https://varg.ai/how-to/video/beat-slideshow.md): Photos cut on the beat of a track, with a title and an end card. - [Music video](https://varg.ai/how-to/video/music-video.md): Scenes cut frame-exact to a song: a beat grid, a story, lyrics and graphics on top. - [Single & album promo](https://varg.ai/how-to/video/release-promo.md): Cover art that moves, a 15-second snippet, the release date and the links. - [Live-coded music video](https://varg.ai/how-to/video/live-code-music.md): Music made in code, shown as code: tokens light on the notes they play. - [Logo sting](https://varg.ai/how-to/video/logo-sting.md): Two to six seconds that end on the logo: build-up, reveal, hold. - [Brand intro & bumper](https://varg.ai/how-to/video/brand-intro.md): The opener every video of a brand shares: mood, mark, tagline. - [Kinetic manifesto](https://varg.ai/how-to/video/manifesto.md): What a brand believes, said in type that moves on the beat, cut with images. - [Title sequence](https://varg.ai/how-to/video/title-sequence.md): Opening credits over a mood: names arrive, the title lands. - [Brand guidelines film](https://varg.ai/how-to/video/brand-guidelines.md): A brand system presented: mark, palette, type, imagery, in one film. - [Stream overlay package](https://varg.ai/how-to/video/stream-overlay.md): Starting soon, live, be right back, ending: loops and overlays for a stream. - [News broadcast package](https://varg.ai/how-to/video/news-package.md): The open, the anchor name, the ticker and the breaking banner a newsroom reuses. - [Countdown & launch timer](https://varg.ai/how-to/video/countdown.md): A clock to zero with something worth waiting for on the other side. - [Conference & event promo](https://varg.ai/how-to/video/event-promo.md): Speakers, agenda, venue and tickets — rendered from the event’s own data. - [Photo memories](https://varg.ai/how-to/video/memories.md): A folder of photos into a film: slow pushes, places on a map, a title and a last line. - [Birthday, wedding & invitation](https://varg.ai/how-to/video/celebration.md): An invite or a greeting: names, photos, date and place, a little confetti. - [Travel route video](https://varg.ai/how-to/video/travel-route.md): A trip retold on a map: the route draws, each stop shows its photos, the totals land. - [Certificate & achievement recap](https://varg.ai/how-to/video/achievement.md): A badge earned, the numbers behind it, a name in lights and a card to share. - [Personalized sales video](https://varg.ai/how-to/video/personalized-outreach.md): One video per prospect: their name, their site in a browser, the pitch that fits. - [Product feed to videos](https://varg.ai/how-to/video/catalog-videos.md): A video for every product in the feed: images, specs, price, CTA — re-rendered when the feed changes. - [Real estate listing](https://varg.ai/how-to/video/real-estate.md): A listing as a tour: exterior, rooms, the specs, the map, the contact. - [Job post video](https://varg.ai/how-to/video/job-post.md): A role, the perks, the team and how to apply — one per open position. - [Recipe video](https://varg.ai/how-to/video/recipe.md): The dish, the ingredients, timed steps, served — from a recipe card. - [Localized versions](https://varg.ai/how-to/video/localized.md): One video in every language: translated captions and type, right-to-left where needed, the same timing. - [Shader loop](https://varg.ai/how-to/video/shader-loop.md): A seamless GPU field — gradients, liquid, light — for backgrounds, screens and stages. - [Generative art loop](https://varg.ai/how-to/video/generative-loop.md): Rules and a seed, rendered frame-exact: flow fields, growth, tilings. - [ASCII & retro looks](https://varg.ai/how-to/video/ascii-retro.md): Footage or type re-rendered as ASCII, CRT, VHS, halftone or pixels. - [3D product spin](https://varg.ai/how-to/video/product-spin.md): A 3D model turning in light, callouts on the parts, the logo at rest. - [Ambient & lofi loop](https://varg.ai/how-to/video/ambient-loop.md): A scene that never ends: weather, particles, grain and a track title. - [Paper collage & zine](https://varg.ai/how-to/video/collage.md): Cut-outs, halftone, torn edges and handwritten notes, moving like a zine. - [Slides to video](https://varg.ai/how-to/video/slides-to-video.md): A deck that plays: each slide builds, charts animate, transitions carry the argument. - [Pitch deck video](https://varg.ai/how-to/video/pitch-video.md): A fundraising story in two minutes: problem, product, traction, team, ask. - [Keynote reveal](https://varg.ai/how-to/video/keynote-reveal.md): The stage moment: a slow build, the product in 3D, the price, the date. - [Game trailer](https://varg.ai/how-to/video/game-trailer.md): Gameplay cut on the music, features in big type, the logo and the date. - [Gameplay highlight & HUD](https://varg.ai/how-to/video/game-overlay.md): Clips from a match with kill feeds, scores and callouts layered on top. ## Prompts — AI video prompts with examples: Seedance, Kling, Veo ### Cinematic & short films - [FLUX 3 Cinematic Inserts](https://varg.ai/how-to/skills/flux-3-cinematic-inserts.md): Write FLUX 3 prompts for standalone cinematic shots: shot craft, camera, continuity and review. Write FLUX 3 prompts for standalone cinematic shots: shot craft, camera, continuity and review. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Cinematic & short films](https://varg.ai/how-to/browse/skills/cinematic.md) - **Slug:** `flux-3-cinematic-inserts` - **Status:** live - **Publisher:** Black Forest Labs (vendor, verified) - **License:** MIT - **Source:** https://github.com/black-forest-labs/skills/tree/master/skills/flux-3-cinematic-inserts - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** BFL API key (BFL_API_KEY) - **Cost:** Billed by Black Forest Labs (API credits) - **Adoption:** 150 installs, 135 GitHub stars (skills.sh, 2026-10-26) ##### What it does Prompting a standalone FLUX 3 text-to-video shot such as an insert or b-roll: shot craft, reliable concept patterns, camera, continuity and review. It ships a field guide, a review scorecard and a JSON brief template. ##### Use when - A single cinematic shot or b-roll insert - Reviewing a generated shot against a scorecard ##### Install - Claude Code: `/plugin marketplace add black-forest-labs/skills && /plugin install flux-3-video@black-forest-labs` - Codex: `npx skills add black-forest-labs/skills --skill flux-3-cinematic-inserts -a codex` - Cursor: `npx skills add black-forest-labs/skills --skill flux-3-cinematic-inserts -a cursor` - Gemini CLI: `npx skills add black-forest-labs/skills --skill flux-3-cinematic-inserts -a gemini-cli` **Models it drives:** flux_3 ##### Example output https://s3.varg.ai/how-to/previews/52479c636774/flux-3-cinematic-inserts.mp4 — FLUX 3 'Volumetric light' example from BFL's video prompting cheatsheet: god rays slicing through a dusty barn, motes drifting — a standalone atmospheric shot; 2-10 s (credit: Black Forest Labs, https://docs.bfl.ai/guides/prompting_video_camera_terms) ##### More cinematic & short films skills - [Previz to Seedance](https://varg.ai/how-to/skills/varg-previz.md): Plan multi-shot AI video before paying: story, anchor sheets, a three.js grey-box previz, then a Seedance v2v pack. - [Higgsfield Generate](https://varg.ai/how-to/skills/higgsfield-generate.md): Generate images, video, 3D and audio across 30+ models, plus ad videos and virality scoring, from one CLI. - [fal Recipes](https://varg.ai/how-to/skills/fal-recipes.md): Follow proven multi-step fal pipelines for commercials, product shots, characters, lip-sync and try-on. - [fal Fan Cam](https://varg.ai/how-to/skills/fal-fan-cam.md): Put someone into a live sports broadcast: a realistic fan-cam cutaway with scoreboard and channel overlays. - [Gemini Omni Flash](https://varg.ai/how-to/skills/gemini-omni-flash-api.md): Generate and edit video with Gemini Omni 1.1 Flash: text-to-video, first/last frames, extensions, edits. - [Melies](https://varg.ai/how-to/skills/melies.md): Generate images, video, movie posters and thumbnails from the terminal with 148 AI actors and style flags. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt ### Images & thumbnails - [Douyin Cover Builder](https://varg.ai/how-to/skills/douyin-cover-builder.md): Writes image-model prompts for Douyin covers: bold title, highlighted keywords, your face bottom right Writes image-model prompts for Douyin covers: bold title, highlighted keywords, your face bottom right - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Images & thumbnails](https://varg.ai/how-to/browse/skills/image.md) - **Slug:** `douyin-cover-builder` - **Status:** beta - **Publisher:** 0x00000003 (community) - **License:** unknown - **Source:** https://clawhub.ai/0x00000003/skills/douyin-cover-builder - **Works in:** OpenClaw - **Cost:** Free — writes text only - **Adoption:** 205 installs, 7 ClawHub stars (ClawHub, 2026-10-27) ##### What it does From a topic, optional key points, the pose or mood you want, and optionally your photo, it writes a final prompt for an image model plus a short art direction note. Rules cover a fixed layout (title on top, you bottom right, one hero metaphor bottom left), keywords in a contrast colour on a slight upward arc, and a background palette chosen by topic. It defaults to strict 3:4 (1080×1440) and gives a pixel check for the generated file. A prompt template only; it does not generate the image itself. ##### Use when - You need a cover or thumbnail prompt for a Douyin video - You want a consistent cover series with your face and a bold title - You want keyword highlighting and a topic-matched colour scheme on a cover ##### Not for - Generating the image without a separate image model ##### Install - OpenClaw: `npx clawhub@latest install douyin-cover-builder` ##### More images & thumbnails skills - [Higgsfield YouTube Thumbnail](https://varg.ai/how-to/skills/higgsfield-youtube-thumbnail.md): Design truthful, high-CTR YouTube thumbnails and vertical covers with up to three consistent faces. - [Slack GIF Creator](https://varg.ai/how-to/skills/anthropic-slack-gif-creator.md): Make animated GIFs that pass Slack's limits: 128×128 emoji or 480×480 message GIFs, validated on export. - [PixiJS GIFs](https://varg.ai/how-to/skills/pixijs-gif.md): Load, play and control animated GIFs inside a PixiJS v8 scene. - [AI Image Generation](https://varg.ai/how-to/skills/inference-sh-ai-image-generation.md): Generate images with 50+ models from one CLI: GPT Image 2.5, FLUX, Gemini, Grok, Seedream, Reve and more. - [Background Removal](https://varg.ai/how-to/skills/inference-sh-background-removal.md): Cut subjects out of photos into clean transparent PNGs with BiRefNet. - [FLUX Image](https://varg.ai/how-to/skills/inference-sh-flux-image.md): Generate and restyle images with FLUX Dev LoRA and FLUX.2 Klein, including custom LoRA styles. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt ### Characters - [Grandma Characters](https://varg.ai/how-to/skills/varg-character-grandma.md): Very old AI characters people remember: age vs behaviour, real wrinkles, a loud outfit and one prop. Very old AI characters people remember: age vs behaviour, real wrinkles, a loud outfit and one prop. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Characters](https://varg.ai/how-to/browse/skills/characters.md) - **Slug:** `varg-character-grandma` - **Status:** live - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills/tree/main/varg-character-grandma - **Works in:** Claude Code, Codex, Cursor, Gemini CLI, OpenClaw - **Requires:** varg API key or varg MCP - **Cost:** ~7 credits per sheet, 6 per scene still — Measured on varg 2026-10-27: Seedream 5 Pro sheet 7, Nano Banana 2.1 sheet or still 6, 5 s video 83 (Kling 4) to 203 (Seedance 2.5 with audio). ##### What it does Granny pages have the best money evidence of any AI archetype (Granny Spills: 2.1M followers and brand deals). The recipe: an exact age, real wrinkles and age spots, one loud piece (leopard housecoat, shiny track suit, pink fur coat), one prop and one behaviour that does not fit the age. Ready characters, the sheet prompt, formats that work and the FTC line. Example: Nonna Rosa. ##### Use when - An AI granny, grandpa, babushka, nonna or abuela - An old character for an ad or an AI influencer page ##### Install - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` - Codex: `npx skills add vargHQ/skills --skill varg-character-grandma -a codex -y` - Cursor: `npx skills add vargHQ/skills --skill varg-character-grandma -a cursor -y` - Gemini CLI: `npx skills add vargHQ/skills --skill varg-character-grandma -a gemini-cli -y` - OpenClaw: `npx clawhub@latest install varg-character-grandma` **Models it drives:** seedream_5_pro, nano_banana_2_1, kling_v4, seedance_2_5 ##### Example output https://s3.varg.ai/how-to/previews/30b8895d05ea/varg-character-grandma.webp — Nonna Rosa character sheet, Seedream 5 Pro, 7 credits (2026-10-27) (credit: varg, https://github.com/vargHQ/skills/tree/main/varg-character-grandma) ##### How-to tutorials that use it - [How to make an AI influencer video](https://varg.ai/how-to/make-an-ai-influencer-video.md): A recurring AI character with one look and one format, posting as a creator or a brand face. ##### More characters skills - [Consistent Characters](https://varg.ai/how-to/skills/varg-consistent-characters.md): Multi-scene ads where the same actors, outfits and location hold in every cut: cards, composites, Kling v3 clips. - [Character Design](https://varg.ai/how-to/skills/varg-character-design.md): How to make AI characters people remember: definition, details, clothing, and a sheet that keeps them consistent. - [Big Guy Characters](https://varg.ai/how-to/skills/varg-character-big-guy.md): Big, heavy characters played straight: size shown with visible cues, a small delicate job, never mocked. - [Cringe Deadpan Characters](https://varg.ai/how-to/skills/varg-character-cringe.md): Bizarre but believable characters people argue about: one wrong outfit, one repeated action, no smile. - [Beautiful Women](https://varg.ai/how-to/skills/varg-character-beautiful-woman.md): Very beautiful AI women who look like real models: casting details, real skin, fashion styling, no false blocks. - [Fitness Characters](https://varg.ai/how-to/skills/varg-character-fitness.md): Fitness coaches and athletes with real muscle, a signature colour and catalog wording that gets sports bras through. - [Big Guy Characters](https://varg.ai/how-to/skills/varg-character-big-guy.md): Big, heavy characters played straight: size shown with visible cues, a small delicate job, never mocked. Big, heavy characters played straight: size shown with visible cues, a small delicate job, never mocked. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Characters](https://varg.ai/how-to/browse/skills/characters.md) - **Slug:** `varg-character-big-guy` - **Status:** live - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills/tree/main/varg-character-big-guy - **Works in:** Claude Code, Codex, Cursor, Gemini CLI, OpenClaw - **Requires:** varg API key or varg MCP - **Cost:** ~7 credits per sheet, 6 per scene still — Measured on varg 2026-10-27: Seedream 5 Pro sheet 7, Nano Banana 2.1 sheet or still 6, 5 s video 83 (Kling 4) to 203 (Seedance 2.5 with audio). ##### What it does Extreme bodies played straight spread (a 280-lb clown lifting cars: 14.6M views on one reel). Numbers alone come out too small: the recipe shows size with visible cues (belly, neck, arms, fit) and puts the contrast in the job and the outfit (a 160 kg florist in a tiny pink apron). Ready characters, sheet tips and what not to joke about. Example: Bruno. ##### Use when - A fat, big, heavy or plus-size character - A gentle giant or a strongman ##### Install - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` - Codex: `npx skills add vargHQ/skills --skill varg-character-big-guy -a codex -y` - Cursor: `npx skills add vargHQ/skills --skill varg-character-big-guy -a cursor -y` - Gemini CLI: `npx skills add vargHQ/skills --skill varg-character-big-guy -a gemini-cli -y` - OpenClaw: `npx clawhub@latest install varg-character-big-guy` **Models it drives:** seedream_5_pro, nano_banana_2_1, kling_v4 ##### Example output https://s3.varg.ai/how-to/previews/a8f54ecd7b02/varg-character-big-guy.webp — Bruno model sheet, Seedream 5 Pro with visible size cues, 7 credits (2026-10-27) (credit: varg, https://github.com/vargHQ/skills/tree/main/varg-character-big-guy) ##### More characters skills - [Consistent Characters](https://varg.ai/how-to/skills/varg-consistent-characters.md): Multi-scene ads where the same actors, outfits and location hold in every cut: cards, composites, Kling v3 clips. - [Character Design](https://varg.ai/how-to/skills/varg-character-design.md): How to make AI characters people remember: definition, details, clothing, and a sheet that keeps them consistent. - [Grandma Characters](https://varg.ai/how-to/skills/varg-character-grandma.md): Very old AI characters people remember: age vs behaviour, real wrinkles, a loud outfit and one prop. - [Cringe Deadpan Characters](https://varg.ai/how-to/skills/varg-character-cringe.md): Bizarre but believable characters people argue about: one wrong outfit, one repeated action, no smile. - [Beautiful Women](https://varg.ai/how-to/skills/varg-character-beautiful-woman.md): Very beautiful AI women who look like real models: casting details, real skin, fashion styling, no false blocks. - [Fitness Characters](https://varg.ai/how-to/skills/varg-character-fitness.md): Fitness coaches and athletes with real muscle, a signature colour and catalog wording that gets sports bras through. - [Cringe Deadpan Characters](https://varg.ai/how-to/skills/varg-character-cringe.md): Bizarre but believable characters people argue about: one wrong outfit, one repeated action, no smile. Bizarre but believable characters people argue about: one wrong outfit, one repeated action, no smile. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Characters](https://varg.ai/how-to/browse/skills/characters.md) - **Slug:** `varg-character-cringe` - **Status:** live - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills/tree/main/varg-character-cringe - **Works in:** Claude Code, Codex, Cursor, Gemini CLI, OpenClaw - **Requires:** varg API key or varg MCP - **Cost:** ~7 credits per sheet, 6 per scene still — Measured on varg 2026-10-27: Seedream 5 Pro sheet 7, Nano Banana 2.1 sheet or still 6, 5 s video 83 (Kling 4) to 203 (Seedance 2.5 with audio). ##### What it does The late-2026 meta: Jean Phil reached 244K followers from 12 posts. The recipe: a real-looking adult with one strange haircut, one strange face feature and a suit nobody wears, completely serious, doing one action in a new ordinary place every post, with motion from footage you own. Ready characters and the posting format. Example: Gérard. ##### Use when - A weird, cringe, awkward or deadpan character - An "is he real?" AI persona like Jean Phil ##### Install - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` - Codex: `npx skills add vargHQ/skills --skill varg-character-cringe -a codex -y` - Cursor: `npx skills add vargHQ/skills --skill varg-character-cringe -a cursor -y` - Gemini CLI: `npx skills add vargHQ/skills --skill varg-character-cringe -a gemini-cli -y` - OpenClaw: `npx clawhub@latest install varg-character-cringe` **Models it drives:** seedream_5_pro, kling_v4, seedance_2_5, minimax_h3_max_recast ##### Example output https://s3.varg.ai/how-to/previews/ea0beb949386/varg-character-cringe.webp — Gérard model sheet, Seedream 5 Pro, 7 credits (2026-10-27) (credit: varg, https://github.com/vargHQ/skills/tree/main/varg-character-cringe) ##### How-to tutorials that use it - [How to make an AI influencer video](https://varg.ai/how-to/make-an-ai-influencer-video.md): A recurring AI character with one look and one format, posting as a creator or a brand face. ##### More characters skills - [Consistent Characters](https://varg.ai/how-to/skills/varg-consistent-characters.md): Multi-scene ads where the same actors, outfits and location hold in every cut: cards, composites, Kling v3 clips. - [Character Design](https://varg.ai/how-to/skills/varg-character-design.md): How to make AI characters people remember: definition, details, clothing, and a sheet that keeps them consistent. - [Grandma Characters](https://varg.ai/how-to/skills/varg-character-grandma.md): Very old AI characters people remember: age vs behaviour, real wrinkles, a loud outfit and one prop. - [Big Guy Characters](https://varg.ai/how-to/skills/varg-character-big-guy.md): Big, heavy characters played straight: size shown with visible cues, a small delicate job, never mocked. - [Beautiful Women](https://varg.ai/how-to/skills/varg-character-beautiful-woman.md): Very beautiful AI women who look like real models: casting details, real skin, fashion styling, no false blocks. - [Fitness Characters](https://varg.ai/how-to/skills/varg-character-fitness.md): Fitness coaches and athletes with real muscle, a signature colour and catalog wording that gets sports bras through. - [Beautiful Women](https://varg.ai/how-to/skills/varg-character-beautiful-woman.md): Very beautiful AI women who look like real models: casting details, real skin, fashion styling, no false blocks. Very beautiful AI women who look like real models: casting details, real skin, fashion styling, no false blocks. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Characters](https://varg.ai/how-to/browse/skills/characters.md) - **Slug:** `varg-character-beautiful-woman` - **Status:** live - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills/tree/main/varg-character-beautiful-woman - **Works in:** Claude Code, Codex, Cursor, Gemini CLI, OpenClaw - **Requires:** varg API key or varg MCP - **Cost:** ~7 credits per sheet, 6 per scene still — Measured on varg 2026-10-27: Seedream 5 Pro sheet 7, Nano Banana 2.1 sheet or still 6, 5 s video 83 (Kling 4) to 203 (Seedance 2.5 with audio). ##### What it does Not the average face and not the plastic "AI girl": cast her like a model agency (striking features one by one), keep real skin, style her (makeup, one jewellery piece, an exactly named outfit) and light it like a shoot. Five tested looks (editorial, red carpet, beauty close-up, party flash, swimwear campaign) and the wording that passes filters. Adults only, no nudity. ##### Use when - A beautiful, gorgeous or attractive woman - An AI model or influencer girl - A glamour, red-carpet or swimwear shot that keeps getting blocked ##### Install - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` - Codex: `npx skills add vargHQ/skills --skill varg-character-beautiful-woman -a codex -y` - Cursor: `npx skills add vargHQ/skills --skill varg-character-beautiful-woman -a cursor -y` - Gemini CLI: `npx skills add vargHQ/skills --skill varg-character-beautiful-woman -a gemini-cli -y` - OpenClaw: `npx clawhub@latest install varg-character-beautiful-woman` **Models it drives:** seedream_5_pro, gpt_image_2_5, nano_banana_2_1, soul ##### Example output https://s3.varg.ai/how-to/previews/6c5966b6ac5e/varg-character-beautiful-woman.webp — Four tested looks: Seedream 5 Pro editorial, GPT Image 2.5 red carpet, Nano Banana 2.1 beauty close-up, Soul party flash (2026-10-27) (credit: varg, https://github.com/vargHQ/skills/tree/main/varg-character-beautiful-woman) ##### How-to tutorials that use it - [How to make an AI influencer video](https://varg.ai/how-to/make-an-ai-influencer-video.md): A recurring AI character with one look and one format, posting as a creator or a brand face. ##### More characters skills - [Consistent Characters](https://varg.ai/how-to/skills/varg-consistent-characters.md): Multi-scene ads where the same actors, outfits and location hold in every cut: cards, composites, Kling v3 clips. - [Character Design](https://varg.ai/how-to/skills/varg-character-design.md): How to make AI characters people remember: definition, details, clothing, and a sheet that keeps them consistent. - [Grandma Characters](https://varg.ai/how-to/skills/varg-character-grandma.md): Very old AI characters people remember: age vs behaviour, real wrinkles, a loud outfit and one prop. - [Big Guy Characters](https://varg.ai/how-to/skills/varg-character-big-guy.md): Big, heavy characters played straight: size shown with visible cues, a small delicate job, never mocked. - [Cringe Deadpan Characters](https://varg.ai/how-to/skills/varg-character-cringe.md): Bizarre but believable characters people argue about: one wrong outfit, one repeated action, no smile. - [Fitness Characters](https://varg.ai/how-to/skills/varg-character-fitness.md): Fitness coaches and athletes with real muscle, a signature colour and catalog wording that gets sports bras through. - [Fitness Characters](https://varg.ai/how-to/skills/varg-character-fitness.md): Fitness coaches and athletes with real muscle, a signature colour and catalog wording that gets sports bras through. Fitness coaches and athletes with real muscle, a signature colour and catalog wording that gets sports bras through. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Characters](https://varg.ai/how-to/browse/skills/characters.md) - **Slug:** `varg-character-fitness` - **Status:** live - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills/tree/main/varg-character-fitness - **Works in:** Claude Code, Codex, Cursor, Gemini CLI, OpenClaw - **Requires:** varg API key or varg MCP - **Cost:** ~7 credits per sheet, 6 per scene still — Measured on varg 2026-10-27: Seedream 5 Pro sheet 7, Nano Banana 2.1 sheet or still 6, 5 s video 83 (Kling 4) to 203 (Seedance 2.5 with audio). ##### What it does Real muscle and skin, one signature colour, clothing named like a product listing and always an activity. The catalog prompt passed on all six image models we tested, while "beautiful girl in a sports bra" was blocked on GPT Image 2.5. Ready characters, the prompt that passes and a swap table. Example: Mara. ##### Use when - A fitness influencer, gym character or trainer for an app - An athlete for an activewear or supplement ad - A character in a sports bra that keeps getting blocked ##### Install - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` - Codex: `npx skills add vargHQ/skills --skill varg-character-fitness -a codex -y` - Cursor: `npx skills add vargHQ/skills --skill varg-character-fitness -a cursor -y` - Gemini CLI: `npx skills add vargHQ/skills --skill varg-character-fitness -a gemini-cli -y` - OpenClaw: `npx clawhub@latest install varg-character-fitness` **Models it drives:** seedream_5_pro, nano_banana_2_1, gpt_image_2_5, kling_v4 ##### Example output https://s3.varg.ai/how-to/previews/970c7b30ba05/varg-character-fitness.webp — Mara model sheet with catalog wording, Seedream 5 Pro, 7 credits (2026-10-27) (credit: varg, https://github.com/vargHQ/skills/tree/main/varg-character-fitness) ##### How-to tutorials that use it - [How to make an AI influencer video](https://varg.ai/how-to/make-an-ai-influencer-video.md): A recurring AI character with one look and one format, posting as a creator or a brand face. - [How to make exercise demo video loops with AI](https://varg.ai/how-to/make-exercise-demo-video-loops.md): Looping exercise demos for a fitness app: one trainer, many exercises, clean background. ##### More characters skills - [Consistent Characters](https://varg.ai/how-to/skills/varg-consistent-characters.md): Multi-scene ads where the same actors, outfits and location hold in every cut: cards, composites, Kling v3 clips. - [Character Design](https://varg.ai/how-to/skills/varg-character-design.md): How to make AI characters people remember: definition, details, clothing, and a sheet that keeps them consistent. - [Grandma Characters](https://varg.ai/how-to/skills/varg-character-grandma.md): Very old AI characters people remember: age vs behaviour, real wrinkles, a loud outfit and one prop. - [Big Guy Characters](https://varg.ai/how-to/skills/varg-character-big-guy.md): Big, heavy characters played straight: size shown with visible cues, a small delicate job, never mocked. - [Cringe Deadpan Characters](https://varg.ai/how-to/skills/varg-character-cringe.md): Bizarre but believable characters people argue about: one wrong outfit, one repeated action, no smile. - [Beautiful Women](https://varg.ai/how-to/skills/varg-character-beautiful-woman.md): Very beautiful AI women who look like real models: casting details, real skin, fashion styling, no false blocks. - [Absurd Characters & Mascots](https://varg.ai/how-to/skills/varg-character-absurd.md): Food or object + animal + one human detail: brainrot-style creatures and brand mascots in 3D or clay. Food or object + animal + one human detail: brainrot-style creatures and brand mascots in 3D or clay. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Characters](https://varg.ai/how-to/browse/skills/characters.md) - **Slug:** `varg-character-absurd` - **Status:** live - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills/tree/main/varg-character-absurd - **Works in:** Claude Code, Codex, Cursor, Gemini CLI, OpenClaw - **Requires:** varg API key or varg MCP - **Cost:** ~6 credits per sheet — Nano Banana 2.1, measured 2026-10-27. ##### What it does The fastest-growing AI characters (Italian brainrot, fruit dramas) and the best brand evidence (mascots). The formula, the 3D family-film and clay stop-motion style blocks, the turnaround sheet and three tested characters: Baguettosaurus, Señor Pincho the cactus and Hector the sheep. Original creatures only. ##### Use when - An absurd, funny or brainrot character - A creature, a talking animal or object - A brand mascot ##### Install - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` - Codex: `npx skills add vargHQ/skills --skill varg-character-absurd -a codex -y` - Cursor: `npx skills add vargHQ/skills --skill varg-character-absurd -a cursor -y` - Gemini CLI: `npx skills add vargHQ/skills --skill varg-character-absurd -a gemini-cli -y` - OpenClaw: `npx clawhub@latest install varg-character-absurd` **Models it drives:** nano_banana_2_1, gpt_image_2_5, seedance_2_5, kling_v4 ##### Example output https://s3.varg.ai/how-to/previews/067ffd2edd78/varg-character-absurd.webp — Baguettosaurus, Señor Pincho and Hector from their Nano Banana 2.1 sheets (2026-10-27) (credit: varg, https://github.com/vargHQ/skills/tree/main/varg-character-absurd) ##### How-to tutorials that use it - [How to keep a character consistent in AI video](https://varg.ai/how-to/keep-a-character-consistent-in-ai-video.md): One mascot, trainer or actor that looks identical across many images and clips. ##### More characters skills - [Consistent Characters](https://varg.ai/how-to/skills/varg-consistent-characters.md): Multi-scene ads where the same actors, outfits and location hold in every cut: cards, composites, Kling v3 clips. - [Character Design](https://varg.ai/how-to/skills/varg-character-design.md): How to make AI characters people remember: definition, details, clothing, and a sheet that keeps them consistent. - [Grandma Characters](https://varg.ai/how-to/skills/varg-character-grandma.md): Very old AI characters people remember: age vs behaviour, real wrinkles, a loud outfit and one prop. - [Big Guy Characters](https://varg.ai/how-to/skills/varg-character-big-guy.md): Big, heavy characters played straight: size shown with visible cues, a small delicate job, never mocked. - [Cringe Deadpan Characters](https://varg.ai/how-to/skills/varg-character-cringe.md): Bizarre but believable characters people argue about: one wrong outfit, one repeated action, no smile. - [Beautiful Women](https://varg.ai/how-to/skills/varg-character-beautiful-woman.md): Very beautiful AI women who look like real models: casting details, real skin, fashion styling, no false blocks. - [Cartoon Look Characters](https://varg.ai/how-to/skills/varg-character-cartoon.md): Original characters in famous cartoon looks (yellow sitcom, 3D family film, anime, clay), described without IP names. Original characters in famous cartoon looks (yellow sitcom, 3D family film, anime, clay), described without IP names. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Characters](https://varg.ai/how-to/browse/skills/characters.md) - **Slug:** `varg-character-cartoon` - **Status:** live - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills/tree/main/varg-character-cartoon - **Works in:** Claude Code, Codex, Cursor, Gemini CLI, OpenClaw - **Requires:** varg API key or varg MCP - **Cost:** ~6 credits per sheet — Nano Banana 2.1, measured 2026-10-27. ##### What it does Describe the look, never name the show: Nano Banana 2.1 refused "The Simpsons style" and drew the same look from a description. Style blocks for nine famous looks, what to keep away from in each, a tested original (Gary Pruitt) and a restyle of an existing character from its sheet. ##### Use when - A Simpsons-, Pixar-, Ghibli- or South Park-style character - A cartoon version of a person or a mascot ##### Install - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` - Codex: `npx skills add vargHQ/skills --skill varg-character-cartoon -a codex -y` - Cursor: `npx skills add vargHQ/skills --skill varg-character-cartoon -a cursor -y` - Gemini CLI: `npx skills add vargHQ/skills --skill varg-character-cartoon -a gemini-cli -y` - OpenClaw: `npx clawhub@latest install varg-character-cartoon` **Models it drives:** nano_banana_2_1, gpt_image_2_5, seedance_2_5, minimax_h3_max_style_retro_toon_70s ##### Example output https://s3.varg.ai/how-to/previews/ed07b62f9de5/varg-character-cartoon.webp — Gary Pruitt model sheet in the yellow 2D sitcom look, described without the show's name, Nano Banana 2.1 (2026-10-27) (credit: varg, https://github.com/vargHQ/skills/tree/main/varg-character-cartoon) ##### How-to tutorials that use it - [How to keep a character consistent in AI video](https://varg.ai/how-to/keep-a-character-consistent-in-ai-video.md): One mascot, trainer or actor that looks identical across many images and clips. ##### More characters skills - [Consistent Characters](https://varg.ai/how-to/skills/varg-consistent-characters.md): Multi-scene ads where the same actors, outfits and location hold in every cut: cards, composites, Kling v3 clips. - [Character Design](https://varg.ai/how-to/skills/varg-character-design.md): How to make AI characters people remember: definition, details, clothing, and a sheet that keeps them consistent. - [Grandma Characters](https://varg.ai/how-to/skills/varg-character-grandma.md): Very old AI characters people remember: age vs behaviour, real wrinkles, a loud outfit and one prop. - [Big Guy Characters](https://varg.ai/how-to/skills/varg-character-big-guy.md): Big, heavy characters played straight: size shown with visible cues, a small delicate job, never mocked. - [Cringe Deadpan Characters](https://varg.ai/how-to/skills/varg-character-cringe.md): Bizarre but believable characters people argue about: one wrong outfit, one repeated action, no smile. - [Beautiful Women](https://varg.ai/how-to/skills/varg-character-beautiful-woman.md): Very beautiful AI women who look like real models: casting details, real skin, fashion styling, no false blocks. - [fal Character Design](https://varg.ai/how-to/skills/fal-character-design.md): Design original characters that stay consistent: reference and expression sheets, outfits, character-to-video. Design original characters that stay consistent: reference and expression sheets, outfits, character-to-video. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Characters](https://varg.ai/how-to/browse/skills/characters.md) - **Slug:** `fal-character-design` - **Status:** live - **Publisher:** fal (vendor, verified) - **License:** MIT - **Source:** https://github.com/fal-ai-community/skills/tree/main/skills/character-design - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** fal API key (FAL_KEY), genmedia CLI - **Cost:** Billed by fal per endpoint call - **Adoption:** 457 installs, 256 GitHub stars (skills.sh, 2026-10-26) ##### What it does Builds consistent character media with genmedia: original characters, reference and expression sheets, outfit variations, identity-preserving edits and character-to-video. An anchor system keeps identity stable across shots, with model routing and a quality bar. ##### Use when - An original character or mascot - Reference, turnaround or expression sheets - Keeping one character consistent into video ##### Install - Claude Code: `npx skills add fal-ai-community/skills --skill character-design -a claude-code` - Codex: `npx skills add fal-ai-community/skills --skill character-design -a codex` - Cursor: `npx skills add fal-ai-community/skills --skill character-design -a cursor` - Gemini CLI: `npx skills add fal-ai-community/skills --skill character-design -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/7d6f54829e6f/fal-character-design.webp — 8-panel character turnaround/reference sheet generated on fal (GPT Image 2.5 edit) in fal's character-consistency guide (credit: fal, https://fal.ai/learn/tools/ai-character-consistency) ##### How-to tutorials that use it - [How to keep a character consistent in AI video](https://varg.ai/how-to/keep-a-character-consistent-in-ai-video.md): One mascot, trainer or actor that looks identical across many images and clips. - [How to make exercise demo video loops with AI](https://varg.ai/how-to/make-exercise-demo-video-loops.md): Looping exercise demos for a fitness app: one trainer, many exercises, clean background. ##### More characters skills - [Consistent Characters](https://varg.ai/how-to/skills/varg-consistent-characters.md): Multi-scene ads where the same actors, outfits and location hold in every cut: cards, composites, Kling v3 clips. - [Character Design](https://varg.ai/how-to/skills/varg-character-design.md): How to make AI characters people remember: definition, details, clothing, and a sheet that keeps them consistent. - [Grandma Characters](https://varg.ai/how-to/skills/varg-character-grandma.md): Very old AI characters people remember: age vs behaviour, real wrinkles, a loud outfit and one prop. - [Big Guy Characters](https://varg.ai/how-to/skills/varg-character-big-guy.md): Big, heavy characters played straight: size shown with visible cues, a small delicate job, never mocked. - [Cringe Deadpan Characters](https://varg.ai/how-to/skills/varg-character-cringe.md): Bizarre but believable characters people argue about: one wrong outfit, one repeated action, no smile. - [Beautiful Women](https://varg.ai/how-to/skills/varg-character-beautiful-woman.md): Very beautiful AI women who look like real models: casting details, real skin, fashion styling, no false blocks. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Character Design Sheet](https://varg.ai/how-to/skills/inference-sh-character-design-sheet.md): Keep a character consistent across AI images with turnarounds, expression sheets, palettes and LoRAs. Keep a character consistent across AI images with turnarounds, expression sheets, palettes and LoRAs. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Characters](https://varg.ai/how-to/browse/skills/characters.md) - **Slug:** `inference-sh-character-design-sheet` - **Status:** live - **Publisher:** inference.sh (vendor, verified) - **License:** MIT - **Source:** https://github.com/inference-sh/skills/tree/main/guides/design/character-design-sheet - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** inference.sh account, belt CLI (`belt login`) - **Cost:** Billed by inference.sh - **Adoption:** 11 installs, 781 GitHub stars (skills.sh, 2026-10-26) ##### What it does Character consistency for AI-generated images: reference sheets with turnaround views and expression sheets, color palettes, style-consistency techniques and LoRA approaches. The sheets are generated with FLUX through inference.sh. ##### Use when - A character reference or turnaround sheet - The same character across many images ##### Install - Claude Code: `npx skills add inference-sh/skills --skill character-design-sheet -a claude-code` - Codex: `npx skills add inference-sh/skills --skill character-design-sheet -a codex` - Cursor: `npx skills add inference-sh/skills --skill character-design-sheet -a cursor` - Gemini CLI: `npx skills add inference-sh/skills --skill character-design-sheet -a gemini-cli` **Models it drives:** flux_dev ##### How-to tutorials that use it - [How to keep a character consistent in AI video](https://varg.ai/how-to/keep-a-character-consistent-in-ai-video.md): One mascot, trainer or actor that looks identical across many images and clips. ##### More characters skills - [Consistent Characters](https://varg.ai/how-to/skills/varg-consistent-characters.md): Multi-scene ads where the same actors, outfits and location hold in every cut: cards, composites, Kling v3 clips. - [Character Design](https://varg.ai/how-to/skills/varg-character-design.md): How to make AI characters people remember: definition, details, clothing, and a sheet that keeps them consistent. - [Grandma Characters](https://varg.ai/how-to/skills/varg-character-grandma.md): Very old AI characters people remember: age vs behaviour, real wrinkles, a loud outfit and one prop. - [Big Guy Characters](https://varg.ai/how-to/skills/varg-character-big-guy.md): Big, heavy characters played straight: size shown with visible cues, a small delicate job, never mocked. - [Cringe Deadpan Characters](https://varg.ai/how-to/skills/varg-character-cringe.md): Bizarre but believable characters people argue about: one wrong outfit, one repeated action, no smile. - [Beautiful Women](https://varg.ai/how-to/skills/varg-character-beautiful-woman.md): Very beautiful AI women who look like real models: casting details, real skin, fashion styling, no false blocks. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt ### Sound effects - [FLUX 3 Audio & Dialogue](https://varg.ai/how-to/skills/flux-3-audio-dialogue.md): Direct FLUX 3's generated sound: dialogue, voiceover, ambience, effects, music, silence and timing. Direct FLUX 3's generated sound: dialogue, voiceover, ambience, effects, music, silence and timing. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Sound effects](https://varg.ai/how-to/browse/skills/sfx.md) - **Slug:** `flux-3-audio-dialogue` - **Status:** live - **Publisher:** Black Forest Labs (vendor, verified) - **License:** MIT - **Source:** https://github.com/black-forest-labs/skills/tree/master/skills/flux-3-audio-dialogue - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** BFL API key (BFL_API_KEY) - **Cost:** Billed by Black Forest Labs (API credits) - **Adoption:** 146 installs, 135 GitHub stars (skills.sh, 2026-10-26) ##### What it does Directs the audio that FLUX 3 generates together with the picture. It covers dialogue and voiceover, ambience, effects, music, silence and timing cues in the prompt. ##### Use when - Dialogue or voiceover in a FLUX 3 clip - Sound design in generated video ##### Install - Claude Code: `/plugin marketplace add black-forest-labs/skills && /plugin install flux-3-video@black-forest-labs` - Codex: `npx skills add black-forest-labs/skills --skill flux-3-audio-dialogue -a codex` - Cursor: `npx skills add black-forest-labs/skills --skill flux-3-audio-dialogue -a cursor` - Gemini CLI: `npx skills add black-forest-labs/skills --skill flux-3-audio-dialogue -a gemini-cli` **Models it drives:** flux_3 ##### Example output https://s3.varg.ai/how-to/previews/095175a46912/flux-3-audio-dialogue.mp3 — Soundtrack of FLUX 3's 'Multilingual dialogue' example (a telenovela scene with a woman in red and two men in white suits): generated dialogue and room tone, audio only (credit: Black Forest Labs, https://bfl.ai/models/flux-3-video) ##### More sound effects skills - [ElevenLabs Sound Effects](https://varg.ai/how-to/skills/elevenlabs-sound-effects.md): Generate sound effects and ambiences from a text description, including seamless loops. - [ElevenLabs Voice Isolator](https://varg.ai/how-to/skills/elevenlabs-voice-isolator.md): Strip background noise and music from dialogue to get clean, isolated speech. - [ElevenLabs Sound Effects (inference.sh)](https://varg.ai/how-to/skills/inference-sh-elevenlabs-sound-effects.md): Generate royalty-free sound effects from text with ElevenLabs, via inference.sh. - [ElevenLabs Voice Isolator (inference.sh)](https://varg.ai/how-to/skills/inference-sh-elevenlabs-voice-isolator.md): Remove background noise and isolate vocals from any recording with ElevenLabs, via inference.sh. - [Kling SFX via dLazy](https://varg.ai/how-to/skills/dlazy-keling-sfx.md): Generate sound effects from a text prompt or matched to a video with Kling SFX, through the dLazy CLI. ### Camera & shots - [fal Cinematography](https://varg.ai/how-to/skills/fal-cinematography.md): Direct cinematic stills and video: shot language, camera moves, lenses, lighting, grade and film texture. Direct cinematic stills and video: shot language, camera moves, lenses, lighting, grade and film texture. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Camera & shots](https://varg.ai/how-to/browse/skills/camera.md) - **Slug:** `fal-cinematography` - **Status:** live - **Publisher:** fal (vendor, verified) - **License:** MIT - **Source:** https://github.com/fal-ai-community/skills/tree/main/skills/cinematography - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** fal API key (FAL_KEY), genmedia CLI - **Cost:** Billed by fal per endpoint call - **Adoption:** 1,100 installs, 256 GitHub stars (skills.sh, 2026-10-26) ##### What it does A production skill for cinematic image and video prompts run through genmedia: it collects the shot intent, builds the prompt in a set order, routes to a model and checks the result against a quality bar. References cover shot language, lighting, lens and color choices, with worked examples. ##### Use when - A cinematic still or shot - Camera movement, lens or lighting direction - A specific grade or film texture ##### Install - Claude Code: `npx skills add fal-ai-community/skills --skill cinematography -a claude-code` - Codex: `npx skills add fal-ai-community/skills --skill cinematography -a codex` - Cursor: `npx skills add fal-ai-community/skills --skill cinematography -a cursor` - Gemini CLI: `npx skills add fal-ai-community/skills --skill cinematography -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/825db19fa1b1/fal-cinematography.mp4 — Seedance 2.0 example on fal's model page: 100mm macro shot with a slow slide and rack focus from tiger fur to the eye; 2-10 s (credit: fal, https://fal.ai/models/bytedance/seedance-2.0/text-to-video) ##### How-to tutorials that use it - [How to make an AI short film](https://varg.ai/how-to/make-an-ai-short-film.md): A multi-shot story with consistent characters: script, storyboard, previz, shots, edit. ##### More camera & shots skills - [Storyboard Creation](https://varg.ai/how-to/skills/inference-sh-storyboard-creation.md): Storyboard a video before generating: shot types, angles, moves, the 180-degree rule, annotated panels. - [Video Shotcraft](https://varg.ai/how-to/skills/video-shotcraft.md): 124 shot recipe cards and a tested Remotion template for cinematic product promos built from real UI screenshots. - [Camera Moves](https://varg.ai/how-to/skills/craft-camera-moves.md): 47 camera moves with direction, speed and an end frame — dolly in to car mount — in six models' syntax. - [Shot Sizes & Angles](https://varg.ai/how-to/skills/craft-shot-sizes-angles.md): 29 framing cards — extreme wide to insert, cowboy to master shot, eye level to worm's-eye — with model-ready clauses. - [Lens & Optics](https://varg.ai/how-to/skills/craft-lens-optics.md): 20 lens and focus cards — anamorphic, rack focus, lens flare, chromatic aberration — written as visible effects. - [Composition & Framing](https://varg.ai/how-to/skills/craft-composition-framing.md): 34 composition cards — thirds, symmetry, leading lines, forced perspective, 2.39:1 to IMAX — for stills and video. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Video Shotcraft](https://varg.ai/how-to/skills/video-shotcraft.md): 124 shot recipe cards and a tested Remotion template for cinematic product promos built from real UI screenshots. 124 shot recipe cards and a tested Remotion template for cinematic product promos built from real UI screenshots. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Camera & shots](https://varg.ai/how-to/browse/skills/camera.md) - **Slug:** `video-shotcraft` - **Status:** live - **Publisher:** Vincent Wei (community) - **License:** Apache-2.0 - **Source:** https://github.com/Vincentwei1021/video-shotcraft - **Works in:** Claude Code, Codex - **Requires:** Node.js + Remotion, ffmpeg, Python 3 (helper scripts), Optional: Upload-Post API key for publishing - **Cost:** Free — local Remotion render - **Adoption:** 5,310 installs, 13,824 GitHub stars (GitHub + skills.sh, 2026-10-26) ##### What it does A self-contained production kit for web and desktop product promos: 124 shot recipe cards in 10 categories (each with intent, a motion core, a parameter table, known pitfalls and a reference implementation), a validated Remotion template called Ink Press, reusable components and audio assets, and a six-stage workflow. The skill first runs a read-only product check, then offers three modes: reuse the template, autonomous free creation, or co-creation with sign-off on the brief, visual direction, shot mapping and storyboard. The cards double as a general motion vocabulary, so you can pull one for a single shot in any video. Extras export Jianying (CapCut China) drafts and publish through Upload-Post. ##### Use when - Turn a web app or frontend project into a cinematic product video - Reuse the Ink Press template for your own product - Apply one named shot card to a single clip ##### Not for - Live-action or AI-generated footage (it animates real page screenshots in code) ##### Install - Claude Code: `npx skills add Vincentwei1021/video-shotcraft` - Codex: `npx skills add Vincentwei1021/video-shotcraft` ##### Example output https://s3.varg.ai/how-to/previews/2debe3ea5ec0/video-shotcraft.mp4 — The exploded-view shot card from the skill's Gallery: a product dashboard tilts in 3D, its panels fly apart with callouts, then reassemble under 'Built in layers.' Full 5.7 s clip. (credit: Vincent Wei (video-shotcraft), https://vincentwei1021.github.io/video-shotcraft/) ##### How-to tutorials that use it - [How to make a SaaS launch video with AI](https://varg.ai/how-to/make-a-saas-launch-video.md): A polished launch film for an app or website: UI cards, device mockups, kinetic type, music. - [How to make an app demo video with AI](https://varg.ai/how-to/make-an-app-demo-video.md): A polished demo of your app: device mockups, real screens, captions, music — for the App Store, ads or a landing page. ##### More camera & shots skills - [fal Cinematography](https://varg.ai/how-to/skills/fal-cinematography.md): Direct cinematic stills and video: shot language, camera moves, lenses, lighting, grade and film texture. - [Storyboard Creation](https://varg.ai/how-to/skills/inference-sh-storyboard-creation.md): Storyboard a video before generating: shot types, angles, moves, the 180-degree rule, annotated panels. - [Camera Moves](https://varg.ai/how-to/skills/craft-camera-moves.md): 47 camera moves with direction, speed and an end frame — dolly in to car mount — in six models' syntax. - [Shot Sizes & Angles](https://varg.ai/how-to/skills/craft-shot-sizes-angles.md): 29 framing cards — extreme wide to insert, cowboy to master shot, eye level to worm's-eye — with model-ready clauses. - [Lens & Optics](https://varg.ai/how-to/skills/craft-lens-optics.md): 20 lens and focus cards — anamorphic, rack focus, lens flare, chromatic aberration — written as visible effects. - [Composition & Framing](https://varg.ai/how-to/skills/craft-composition-framing.md): 34 composition cards — thirds, symmetry, leading lines, forced perspective, 2.39:1 to IMAX — for stills and video. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Camera Moves](https://varg.ai/how-to/skills/craft-camera-moves.md): 47 camera moves with direction, speed and an end frame — dolly in to car mount — in six models' syntax. 47 camera moves with direction, speed and an end frame — dolly in to car mount — in six models' syntax. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Camera & shots](https://varg.ai/how-to/browse/skills/camera.md) - **Slug:** `craft-camera-moves` - **Status:** beta - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills - **Works in:** Claude Code, Claude.ai, Codex, Cursor, Gemini CLI, OpenClaw, Any MCP client, varg Studio ##### What it does 47 original camera-move cards, from dolly in and bullet time to Steadicam glides, vehicle and car mounts, hyperlapse and found-footage handheld. Every clause names one move, its direction and speed, what it moves relative to, and the frame it settles on — the pattern that stops the drifting, never-settling camera. Aliases map other tools' vocabulary onto each card (Higgsfield's 'Super Dolly In' and 'Lazy Susan', Hailuo's `[Truck left]`, Luma's 'Orbit Right', Seedance's 环绕 and 希区柯克变焦), and adapters cover Veo, Kling, Seedance, Hailuo, Luma and Runway. Moves Hailuo has no single command for are composed from up to three simultaneous commands — an orbit becomes `[Truck left,Pan right]`, a crane becomes `[Pedestal up,Tilt down]`. ##### Use when - Choosing or naming a camera move for a shot - Translating a preset name from another app into a move any model understands - A generated clip drifts or never settles — rewrite the move with an explicit end frame ##### Not for - Lens and focus effects (craft-lens-optics) - Several moves stacked in one clip — split them into shots ##### Install Not published as a standalone skill yet. Source: https://github.com/vargHQ/skills **Models it drives:** veo_3_1, veo_3_1_fast, veo_3_1_lite, omni_flash_1_1, kling_v3, kling_v3_standard, seedance_2_5, seedance_2, minimax_h3_max ##### Example output https://s3.varg.ai/how-to/previews/b6912b876f91/craft-camera-moves.mp4 — A live-action arc: the camera orbits Thom as he walks onto a canal bridge and stops, swinging from a 3/4-front view through profile to 3/4-back while the background wheels from the church to the canal. Tears of Steel (2012) 04:30.5-04:38.5, 720p release from the Blender mirror (download.blender.org/demo/movies/ToS), center-cropped by varg from 2.40:1 to 16:10, no audio. (credit: Tears of Steel (CC BY 3.0) — (CC) Blender Foundation | mango.blender.org, https://mango.blender.org/) ##### How-to tutorials that use it - [How to turn a product photo into a video with AI](https://varg.ai/how-to/turn-a-product-photo-into-a-video.md): A product shot turned into a short promo: orbit, push-in, exploded view or bullet time, with music. - [How to animate a photo into a video with AI](https://varg.ai/how-to/animate-a-photo-into-a-video.md): A still photo brought to life: a subtle smile, blink, camera move or full scene motion. - [How to make a property listing video with AI](https://varg.ai/how-to/make-a-property-listing-video.md): A listing reel from property photos: gentle camera motion through rooms, captions with price and features, music. ##### More camera & shots skills - [fal Cinematography](https://varg.ai/how-to/skills/fal-cinematography.md): Direct cinematic stills and video: shot language, camera moves, lenses, lighting, grade and film texture. - [Storyboard Creation](https://varg.ai/how-to/skills/inference-sh-storyboard-creation.md): Storyboard a video before generating: shot types, angles, moves, the 180-degree rule, annotated panels. - [Video Shotcraft](https://varg.ai/how-to/skills/video-shotcraft.md): 124 shot recipe cards and a tested Remotion template for cinematic product promos built from real UI screenshots. - [Shot Sizes & Angles](https://varg.ai/how-to/skills/craft-shot-sizes-angles.md): 29 framing cards — extreme wide to insert, cowboy to master shot, eye level to worm's-eye — with model-ready clauses. - [Lens & Optics](https://varg.ai/how-to/skills/craft-lens-optics.md): 20 lens and focus cards — anamorphic, rack focus, lens flare, chromatic aberration — written as visible effects. - [Composition & Framing](https://varg.ai/how-to/skills/craft-composition-framing.md): 34 composition cards — thirds, symmetry, leading lines, forced perspective, 2.39:1 to IMAX — for stills and video. - [Shot Sizes & Angles](https://varg.ai/how-to/skills/craft-shot-sizes-angles.md): 29 framing cards — extreme wide to insert, cowboy to master shot, eye level to worm's-eye — with model-ready clauses. 29 framing cards — extreme wide to insert, cowboy to master shot, eye level to worm's-eye — with model-ready clauses. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Camera & shots](https://varg.ai/how-to/browse/skills/camera.md) - **Slug:** `craft-shot-sizes-angles` - **Status:** beta - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills - **Works in:** Claude Code, Claude.ai, Codex, Cursor, Gemini CLI, OpenClaw, Any MCP client, varg Studio ##### What it does 12 shot sizes and 17 camera angles as original cards: what each framing shows, when a director reaches for it, and the near-miss that reads differently (a high angle is not an overhead; a worm's-eye is not a low angle). Aliases cover crew shorthand (EWS, MCU, OTS), Kling's shot labels ('Medium Close up', 'Full Body Shot'), Luma's nine angle concepts and Seedance's 特写 / 近景 / 中景 / 全景 / 大全景. Deep links send humans to shot.cafe tag pages and eyecannndy technique pages; the agent never fetches them. ##### Use when - Deciding how tight to frame a subject or where to put the camera - Writing shot labels for a multi-shot prompt - Matching framing across clips from different models ##### Not for - Camera movement (craft-camera-moves) - Lens choice and depth of field (craft-lens-optics) ##### Install Not published as a standalone skill yet. Source: https://github.com/vargHQ/skills **Models it drives:** veo_3_1, veo_3_1_fast, veo_3_1_lite, omni_flash_1_1, kling_v3, kling_v3_standard, seedance_2_5, seedance_2, minimax_h3_max, nano_banana_2_1, flux_3_image ##### Example output https://s3.varg.ai/how-to/previews/6c9ff2fd073f/craft-shot-sizes-angles.mp4 — Two low-angle shots as cut in the film: a worm's-eye view from the grass as the rabbit stretches up into the sky, then a low-angle medium close-up (MCU) of him smiling against the sky. Big Buck Bunny 4K release (Wikimedia Commons 1080p transcode) 01:01.4-01:08.2 (00:57.0-01:03.8 in the 2008 cut), no audio. (credit: Big Buck Bunny (CC BY 3.0) — (c) copyright 2008, Blender Foundation | bigbuckbunny.org, https://commons.wikimedia.org/wiki/File:Big_Buck_Bunny_4K.webm) ##### More camera & shots skills - [fal Cinematography](https://varg.ai/how-to/skills/fal-cinematography.md): Direct cinematic stills and video: shot language, camera moves, lenses, lighting, grade and film texture. - [Storyboard Creation](https://varg.ai/how-to/skills/inference-sh-storyboard-creation.md): Storyboard a video before generating: shot types, angles, moves, the 180-degree rule, annotated panels. - [Video Shotcraft](https://varg.ai/how-to/skills/video-shotcraft.md): 124 shot recipe cards and a tested Remotion template for cinematic product promos built from real UI screenshots. - [Camera Moves](https://varg.ai/how-to/skills/craft-camera-moves.md): 47 camera moves with direction, speed and an end frame — dolly in to car mount — in six models' syntax. - [Lens & Optics](https://varg.ai/how-to/skills/craft-lens-optics.md): 20 lens and focus cards — anamorphic, rack focus, lens flare, chromatic aberration — written as visible effects. - [Composition & Framing](https://varg.ai/how-to/skills/craft-composition-framing.md): 34 composition cards — thirds, symmetry, leading lines, forced perspective, 2.39:1 to IMAX — for stills and video. - [Lens & Optics](https://varg.ai/how-to/skills/craft-lens-optics.md): 20 lens and focus cards — anamorphic, rack focus, lens flare, chromatic aberration — written as visible effects. 20 lens and focus cards — anamorphic, rack focus, lens flare, chromatic aberration — written as visible effects. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Camera & shots](https://varg.ai/how-to/browse/skills/camera.md) - **Slug:** `craft-lens-optics` - **Status:** beta - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills - **Works in:** Claude Code, Claude.ai, Codex, Cursor, Gemini CLI, OpenClaw, Any MCP client, varg Studio ##### What it does Lens language written as effects first, with millimeters optional: most video-model guides document qualitative words (wide-angle, telephoto, shallow depth of field), and of the current guides only Runway's uses focal lengths. The cards cover wide-angle, telephoto compression, macro, fisheye, anamorphic, shallow and deep focus, rack focus (Seedance: 移焦 plus a plain description), tilt-shift and split diopter. Misuse notes flag the classic confusions: a digital crop does not compress, and blur everywhere is not shallow focus. Newer cards add ultra-wide and normal lenses, vintage cine glass, lens flare, chromatic aberration, vignette, refraction, kaleidoscope and 360 cameras. ##### Use when - User asks for a lens look ('85mm', 'anamorphic', 'bokeh', 'everything sharp') - Separating a subject from a busy background, or keeping two planes in focus ##### Not for - Color and film-stock looks (craft-color-looks) - Camera moves such as zoom or dolly zoom (craft-camera-moves) ##### Install Not published as a standalone skill yet. Source: https://github.com/vargHQ/skills **Models it drives:** veo_3_1, veo_3_1_fast, veo_3_1_lite, omni_flash_1_1, kling_v3, kling_v3_standard, seedance_2_5, seedance_2, minimax_h3_max, nano_banana_2_1, flux_3_image ##### Example output https://s3.varg.ai/how-to/previews/b3bc9499ee00/craft-lens-optics.mp4 — Long-lens shallow focus on real glass: two live-action close-ups (Thom, then Celia) with the canal houses and trees behind them melted into soft bokeh so only the face is sharp, the look the '85mm, f/1.4' card describes. Tears of Steel (2012) 05:15.5-05:22.1, 720p release from the Blender mirror (download.blender.org/demo/movies/ToS), full 2.40:1 frame, no audio. (credit: Tears of Steel (CC BY 3.0) — (CC) Blender Foundation | mango.blender.org, https://mango.blender.org/) ##### More camera & shots skills - [fal Cinematography](https://varg.ai/how-to/skills/fal-cinematography.md): Direct cinematic stills and video: shot language, camera moves, lenses, lighting, grade and film texture. - [Storyboard Creation](https://varg.ai/how-to/skills/inference-sh-storyboard-creation.md): Storyboard a video before generating: shot types, angles, moves, the 180-degree rule, annotated panels. - [Video Shotcraft](https://varg.ai/how-to/skills/video-shotcraft.md): 124 shot recipe cards and a tested Remotion template for cinematic product promos built from real UI screenshots. - [Camera Moves](https://varg.ai/how-to/skills/craft-camera-moves.md): 47 camera moves with direction, speed and an end frame — dolly in to car mount — in six models' syntax. - [Shot Sizes & Angles](https://varg.ai/how-to/skills/craft-shot-sizes-angles.md): 29 framing cards — extreme wide to insert, cowboy to master shot, eye level to worm's-eye — with model-ready clauses. - [Composition & Framing](https://varg.ai/how-to/skills/craft-composition-framing.md): 34 composition cards — thirds, symmetry, leading lines, forced perspective, 2.39:1 to IMAX — for stills and video. - [Composition & Framing](https://varg.ai/how-to/skills/craft-composition-framing.md): 34 composition cards — thirds, symmetry, leading lines, forced perspective, 2.39:1 to IMAX — for stills and video. 34 composition cards — thirds, symmetry, leading lines, forced perspective, 2.39:1 to IMAX — for stills and video. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Camera & shots](https://varg.ai/how-to/browse/skills/camera.md) - **Slug:** `craft-composition-framing` - **Status:** beta - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills - **Works in:** Claude Code, Claude.ai, Codex, Cursor, Gemini CLI, OpenClaw, Any MCP client, varg Studio ##### What it does Composition written as placements a model can follow: where the subject sits, where it looks, and what fills the rest of the frame. Covers rule of thirds, centered symmetry, leading lines, frame within a frame, negative space (copy space for ad titles), foreground layering, two-shots, lead room and widescreen 2.39:1. The same clauses drive image models, so a Nano Banana or FLUX first frame can be composed before any video credits are spent. Newer cards add group and interview framings, headroom, golden ratio, diagonal and triangular layouts, forced perspective, reflections, tableau, figure-ground and aspect-ratio switches. ##### Use when - Composing a first frame, a thumbnail or a product hero - Leaving room for titles and captions ##### Not for - Camera angle and height (craft-shot-sizes-angles) ##### Install Not published as a standalone skill yet. Source: https://github.com/vargHQ/skills **Models it drives:** veo_3_1, veo_3_1_fast, veo_3_1_lite, omni_flash_1_1, kling_v3, kling_v3_standard, seedance_2_5, seedance_2, minimax_h3_max, nano_banana_2_1, flux_3_image, gpt_image_2_5 ##### Example output https://s3.varg.ai/how-to/previews/154762caf94b/craft-composition-framing.webp — Rule of thirds in a real frame: Sintel sits on the right-third line with her head near the upper-right intersection, the cliff fills the lower two thirds and the haze leaves negative space for a title. The thin thirds grid was drawn on by varg. Sintel (2010) frame at 11:34, 2048-px release on archive.org, full 2.35:1 frame. (credit: Sintel (CC BY 3.0) — © copyright Blender Foundation | durian.blender.org, https://archive.org/details/Sintel) ##### More camera & shots skills - [fal Cinematography](https://varg.ai/how-to/skills/fal-cinematography.md): Direct cinematic stills and video: shot language, camera moves, lenses, lighting, grade and film texture. - [Storyboard Creation](https://varg.ai/how-to/skills/inference-sh-storyboard-creation.md): Storyboard a video before generating: shot types, angles, moves, the 180-degree rule, annotated panels. - [Video Shotcraft](https://varg.ai/how-to/skills/video-shotcraft.md): 124 shot recipe cards and a tested Remotion template for cinematic product promos built from real UI screenshots. - [Camera Moves](https://varg.ai/how-to/skills/craft-camera-moves.md): 47 camera moves with direction, speed and an end frame — dolly in to car mount — in six models' syntax. - [Shot Sizes & Angles](https://varg.ai/how-to/skills/craft-shot-sizes-angles.md): 29 framing cards — extreme wide to insert, cowboy to master shot, eye level to worm's-eye — with model-ready clauses. - [Lens & Optics](https://varg.ai/how-to/skills/craft-lens-optics.md): 20 lens and focus cards — anamorphic, rack focus, lens flare, chromatic aberration — written as visible effects. ### Lighting - [Lighting Setups](https://varg.ai/how-to/skills/craft-lighting-setups.md): 56 lighting setups — golden hour to chiaroscuro, rain and fog — each with a named source, direction and quality. 56 lighting setups — golden hour to chiaroscuro, rain and fog — each with a named source, direction and quality. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Lighting](https://varg.ai/how-to/browse/skills/lighting.md) - **Slug:** `craft-lighting-setups` - **Status:** beta - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills - **Works in:** Claude Code, Claude.ai, Codex, Cursor, Gemini CLI, OpenClaw, Any MCP client, varg Studio ##### What it does 56 lighting cards that always name a source and a direction instead of a mood word: golden hour, blue hour, low-key, high-key, Rembrandt, hard side light, rim light, silhouette, soft window light, practicals and volumetric light. Aliases include Higgsfield Cinema Studio presets (Silhouette, Contre-jour, Practicals, Window) and the terms Google, Kling and Wan document ('rembrandt lighting', 'subtle god rays', 'rim light'). Light belongs in the generation prompt; pair with craft-color-looks for the grade. Newer cards add portrait setups (three-point, butterfly, loop, short and broad, kicker, hair and eye light), motivated and available light, neon, firelight, gels and film-noir shadows, and weather and atmosphere: rain, fog, snow, smoke, steam, dust and sparks. ##### Use when - Setting the mood of a shot through light - A generated face looks flat or the subject merges with the background ##### Not for - Relighting an existing clip (use a relight model) - Color grading after generation (craft-color-looks) ##### Install Not published as a standalone skill yet. Source: https://github.com/vargHQ/skills **Models it drives:** veo_3_1, veo_3_1_fast, veo_3_1_lite, omni_flash_1_1, kling_v3, kling_v3_standard, seedance_2_5, seedance_2, minimax_h3_max, nano_banana_2_1, flux_3_image ##### Example output https://s3.varg.ai/how-to/previews/b612fefdcb9d/craft-lighting-setups.mp4 — Rim light from a hard backlight: Proog steps into a blown-out doorway and the light behind him wraps his head, shoulders and hands in a glowing edge against the dark curtains, with only a soft fill on his face. Elephants Dream (2006) 01:10.6-01:18.6, Wikimedia Commons 1080p transcode, no audio. (credit: Elephants Dream (CC BY 2.5) — © 2006 Blender Foundation / Netherlands Media Art Institute | elephantsdream.org, https://commons.wikimedia.org/wiki/File:Elephants_Dream_(2006).webm) ### Prompts by model - [fal Prompting](https://varg.ai/how-to/skills/fal-prompting.md): Prompt specific model families well: Kling, GPT Image 2 and Happy Horse guides, loaded only when needed. Prompt specific model families well: Kling, GPT Image 2 and Happy Horse guides, loaded only when needed. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Prompts by model](https://varg.ai/how-to/browse/skills/prompting.md) - **Slug:** `fal-prompting` - **Status:** live - **Publisher:** fal (vendor, verified) - **License:** MIT - **Source:** https://github.com/fal-ai-community/skills/tree/main/skills/fal-prompting - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** fal API key (FAL_KEY), genmedia CLI - **Cost:** Billed by fal per endpoint call - **Adoption:** 513 installs, 256 GitHub stars (skills.sh, 2026-10-26) ##### What it does Model-family prompt craft for fal endpoints, with a table of which reference to load: separate guides for Kling, GPT Image 2 and Happy Horse. Its universal principles include preferring concrete visual facts over prestige adjectives and inspecting an endpoint's schema before assuming a control exists. ##### Use when - How do I prompt Kling or GPT Image 2? - Prompts keep coming back generic - Moving a prompt between model families ##### Install - Claude Code: `npx skills add fal-ai-community/skills --skill fal-prompting -a claude-code` - Codex: `npx skills add fal-ai-community/skills --skill fal-prompting -a codex` - Cursor: `npx skills add fal-ai-community/skills --skill fal-prompting -a cursor` - Gemini CLI: `npx skills add fal-ai-community/skills --skill fal-prompting -a gemini-cli` **Models it drives:** kling_v3, gpt_image_2, happy_horse ##### Example output https://s3.varg.ai/how-to/previews/1afc7b4b4ed0/fal-prompting.mp4 — 'Short legible text' pattern from fal's Happy Horse prompting guide: slow push-in on a bookshop window with a readable 'THE STARS BELOW' cover (credit: fal, https://fal.ai/learn/tools/prompting-happy-horse) ##### More prompts by model skills - [Replicate Prompt Images](https://varg.ai/how-to/skills/replicate-prompt-images.md): Write better image prompts and edits: photographic language, lighting, text rendering, consistent characters. - [Replicate Prompt Videos](https://varg.ai/how-to/skills/replicate-prompt-videos.md): Prompt video models like a director: shots, camera motion, four audio layers, dialogue, time-coded beats. - [Video Prompting Guide](https://varg.ai/how-to/skills/inference-sh-video-prompting-guide.md): Write prompts that video models follow: shot types, camera moves and per-model tips from Veo to Sora. - [FLUX 3 Prompt Doctor](https://varg.ai/how-to/skills/flux-3-prompt-doctor.md): Diagnose a FLUX 3 video brief before spending: resolve missing, conflicting or schema-changing asks. - [FLUX Image Best Practices](https://varg.ai/how-to/skills/flux-image-best-practices.md): Prompt FLUX image models well: structure, JSON prompts, hex colors, typography and multi-reference edits. - [AI Film Director (Visual Skills)](https://varg.ai/how-to/skills/visual-skills-video-director.md): AI film director for video prompts: dramaturgy first (Murch, blocking, montage), then exact syntax per model. - [Replicate Prompt Images](https://varg.ai/how-to/skills/replicate-prompt-images.md): Write better image prompts and edits: photographic language, lighting, text rendering, consistent characters. Write better image prompts and edits: photographic language, lighting, text rendering, consistent characters. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Prompts by model](https://varg.ai/how-to/browse/skills/prompting.md) - **Slug:** `replicate-prompt-images` - **Status:** live - **Publisher:** Replicate (vendor, verified) - **License:** Apache-2.0 - **Source:** https://github.com/replicate/skills/tree/main/skills/prompt-images - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** Replicate account and API token - **Cost:** Billed by Replicate per prediction - **Adoption:** 2,700 installs, 142 GitHub stars (skills.sh, 2026-10-26) ##### What it does Prompting techniques for image generation and editing models on Replicate. It starts by choosing a model through the API rather than from memory, then covers natural-language prompts, camera and lens language, lighting setups, composition, text rendering, style transfer, character consistency and edits such as object removal, background changes, perspective shifts and inpainting. ##### Use when - Writing prompts for an image model - Editing an image with a model - Building an image-generation feature ##### Install - Claude Code: `npx skills add replicate/skills --skill prompt-images -a claude-code` - Codex: `npx skills add replicate/skills --skill prompt-images -a codex` - Cursor: `npx skills add replicate/skills --skill prompt-images -a cursor` - Gemini CLI: `npx skills add replicate/skills --skill prompt-images -a gemini-cli` **Models it drives:** nano_banana_pro, flux-2, seedream-5, imagen-4, ideogram, recraft ##### Example output https://s3.varg.ai/how-to/previews/3659e94b3693/replicate-prompt-images.webp — Nano Banana Pro output from Replicate's 'How to prompt Nano Banana Pro' guide: a comic-style New York weather report with legible lettering, from its world-knowledge prompting section (credit: Replicate, https://replicate.com/blog/how-to-prompt-nano-banana-pro) ##### More prompts by model skills - [fal Prompting](https://varg.ai/how-to/skills/fal-prompting.md): Prompt specific model families well: Kling, GPT Image 2 and Happy Horse guides, loaded only when needed. - [Replicate Prompt Videos](https://varg.ai/how-to/skills/replicate-prompt-videos.md): Prompt video models like a director: shots, camera motion, four audio layers, dialogue, time-coded beats. - [Video Prompting Guide](https://varg.ai/how-to/skills/inference-sh-video-prompting-guide.md): Write prompts that video models follow: shot types, camera moves and per-model tips from Veo to Sora. - [FLUX 3 Prompt Doctor](https://varg.ai/how-to/skills/flux-3-prompt-doctor.md): Diagnose a FLUX 3 video brief before spending: resolve missing, conflicting or schema-changing asks. - [FLUX Image Best Practices](https://varg.ai/how-to/skills/flux-image-best-practices.md): Prompt FLUX image models well: structure, JSON prompts, hex colors, typography and multi-reference edits. - [AI Film Director (Visual Skills)](https://varg.ai/how-to/skills/visual-skills-video-director.md): AI film director for video prompts: dramaturgy first (Murch, blocking, montage), then exact syntax per model. - [Replicate Prompt Videos](https://varg.ai/how-to/skills/replicate-prompt-videos.md): Prompt video models like a director: shots, camera motion, four audio layers, dialogue, time-coded beats. Prompt video models like a director: shots, camera motion, four audio layers, dialogue, time-coded beats. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Prompts by model](https://varg.ai/how-to/browse/skills/prompting.md) - **Slug:** `replicate-prompt-videos` - **Status:** live - **Publisher:** Replicate (vendor, verified) - **License:** Apache-2.0 - **Source:** https://github.com/replicate/skills/tree/main/skills/prompt-videos - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** Replicate account and API token - **Cost:** Billed by Replicate per prediction - **Adoption:** 2,500 installs, 142 GitHub stars (skills.sh, 2026-10-26) ##### What it does Prompting techniques for video generation models on Replicate: layered scene description, shot types, camera motion and position, lens language, an escalation pattern, prompting all four audio layers, dialogue syntax that avoids burned-in subtitles, pronunciation, and multi-shot or time-coded prompts. Like its image sibling, it picks the model through the API first. ##### Use when - Writing prompts for a video model - Dialogue or sound in generated video - Multi-shot or time-coded clips ##### Install - Claude Code: `npx skills add replicate/skills --skill prompt-videos -a claude-code` - Codex: `npx skills add replicate/skills --skill prompt-videos -a codex` - Cursor: `npx skills add replicate/skills --skill prompt-videos -a cursor` - Gemini CLI: `npx skills add replicate/skills --skill prompt-videos -a gemini-cli` **Models it drives:** veo_3_1, seedance_2, wan-2.2 ##### Example output https://s3.varg.ai/how-to/previews/ba56e406f770/replicate-prompt-videos.mp4 — 'Detailed prompt' result from Replicate's 'How to prompt Veo 3' guide: a directed noir scene (a man answers a ringing payphone in a green-lit alley, cutting to close-ups); audio not kept (credit: Replicate, https://replicate.com/blog/using-and-prompting-veo-3) ##### How-to tutorials that use it - [How to animate a photo into a video with AI](https://varg.ai/how-to/animate-a-photo-into-a-video.md): A still photo brought to life: a subtle smile, blink, camera move or full scene motion. ##### More prompts by model skills - [fal Prompting](https://varg.ai/how-to/skills/fal-prompting.md): Prompt specific model families well: Kling, GPT Image 2 and Happy Horse guides, loaded only when needed. - [Replicate Prompt Images](https://varg.ai/how-to/skills/replicate-prompt-images.md): Write better image prompts and edits: photographic language, lighting, text rendering, consistent characters. - [Video Prompting Guide](https://varg.ai/how-to/skills/inference-sh-video-prompting-guide.md): Write prompts that video models follow: shot types, camera moves and per-model tips from Veo to Sora. - [FLUX 3 Prompt Doctor](https://varg.ai/how-to/skills/flux-3-prompt-doctor.md): Diagnose a FLUX 3 video brief before spending: resolve missing, conflicting or schema-changing asks. - [FLUX Image Best Practices](https://varg.ai/how-to/skills/flux-image-best-practices.md): Prompt FLUX image models well: structure, JSON prompts, hex colors, typography and multi-reference edits. - [AI Film Director (Visual Skills)](https://varg.ai/how-to/skills/visual-skills-video-director.md): AI film director for video prompts: dramaturgy first (Murch, blocking, montage), then exact syntax per model. - [Video Prompting Guide](https://varg.ai/how-to/skills/inference-sh-video-prompting-guide.md): Write prompts that video models follow: shot types, camera moves and per-model tips from Veo to Sora. Write prompts that video models follow: shot types, camera moves and per-model tips from Veo to Sora. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Prompts by model](https://varg.ai/how-to/browse/skills/prompting.md) - **Slug:** `inference-sh-video-prompting-guide` - **Status:** live - **Publisher:** inference.sh (vendor, verified) - **License:** MIT - **Source:** https://github.com/inference-sh/skills/tree/main/guides/prompting/video-prompting-guide - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** inference.sh account, belt CLI (`belt login`) - **Cost:** Billed by inference.sh - **Adoption:** 6 installs, 781 GitHub stars (skills.sh, 2026-10-26) ##### What it does Techniques for AI video prompts across Veo, Seedance, Wan, Grok, Kling, Runway, Pika and Sora: shot types, camera movement, and subject and action description. It adds model-specific strategies with examples that run on inference.sh. ##### Use when - Writing a video prompt - Porting a prompt between video models ##### Install - Claude Code: `npx skills add inference-sh/skills --skill video-prompting-guide -a claude-code` - Codex: `npx skills add inference-sh/skills --skill video-prompting-guide -a codex` - Cursor: `npx skills add inference-sh/skills --skill video-prompting-guide -a cursor` - Gemini CLI: `npx skills add inference-sh/skills --skill video-prompting-guide -a gemini-cli` **Models it drives:** veo_3_1, seedance_2, wan_2_5, grok_imagine ##### More prompts by model skills - [fal Prompting](https://varg.ai/how-to/skills/fal-prompting.md): Prompt specific model families well: Kling, GPT Image 2 and Happy Horse guides, loaded only when needed. - [Replicate Prompt Images](https://varg.ai/how-to/skills/replicate-prompt-images.md): Write better image prompts and edits: photographic language, lighting, text rendering, consistent characters. - [Replicate Prompt Videos](https://varg.ai/how-to/skills/replicate-prompt-videos.md): Prompt video models like a director: shots, camera motion, four audio layers, dialogue, time-coded beats. - [FLUX 3 Prompt Doctor](https://varg.ai/how-to/skills/flux-3-prompt-doctor.md): Diagnose a FLUX 3 video brief before spending: resolve missing, conflicting or schema-changing asks. - [FLUX Image Best Practices](https://varg.ai/how-to/skills/flux-image-best-practices.md): Prompt FLUX image models well: structure, JSON prompts, hex colors, typography and multi-reference edits. - [AI Film Director (Visual Skills)](https://varg.ai/how-to/skills/visual-skills-video-director.md): AI film director for video prompts: dramaturgy first (Murch, blocking, montage), then exact syntax per model. - [FLUX 3 Prompt Doctor](https://varg.ai/how-to/skills/flux-3-prompt-doctor.md): Diagnose a FLUX 3 video brief before spending: resolve missing, conflicting or schema-changing asks. Diagnose a FLUX 3 video brief before spending: resolve missing, conflicting or schema-changing asks. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Prompts by model](https://varg.ai/how-to/browse/skills/prompting.md) - **Slug:** `flux-3-prompt-doctor` - **Status:** live - **Publisher:** Black Forest Labs (vendor, verified) - **License:** MIT - **Source:** https://github.com/black-forest-labs/skills/tree/master/skills/flux-3-prompt-doctor - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** BFL API key (BFL_API_KEY) - **Cost:** Billed by Black Forest Labs (API credits) - **Adoption:** 161 installs, 135 GitHub stars (skills.sh, 2026-10-26) ##### What it does Checks a FLUX 3 brief before generation, routing by what must survive and testing feasibility. It resolves missing, conflicting or schema-changing requirements and returns a verdict before any money is spent. ##### Use when - A vague or contradictory video brief - Checking feasibility before generating ##### Install - Claude Code: `/plugin marketplace add black-forest-labs/skills && /plugin install flux-3-video@black-forest-labs` - Codex: `npx skills add black-forest-labs/skills --skill flux-3-prompt-doctor -a codex` - Cursor: `npx skills add black-forest-labs/skills --skill flux-3-prompt-doctor -a cursor` - Gemini CLI: `npx skills add black-forest-labs/skills --skill flux-3-prompt-doctor -a gemini-cli` **Models it drives:** flux_3 ##### Example output https://s3.varg.ai/how-to/previews/0af8e1439181/flux-3-prompt-doctor.mp4 — FLUX 3 text-to-video example from BFL's docs (a beekeeper opening a hive at golden hour) — the kind of video job the prompt doctor checks before spending; 0-8 s (credit: Black Forest Labs, https://docs.bfl.ai/flux_3/flux3_video) ##### More prompts by model skills - [fal Prompting](https://varg.ai/how-to/skills/fal-prompting.md): Prompt specific model families well: Kling, GPT Image 2 and Happy Horse guides, loaded only when needed. - [Replicate Prompt Images](https://varg.ai/how-to/skills/replicate-prompt-images.md): Write better image prompts and edits: photographic language, lighting, text rendering, consistent characters. - [Replicate Prompt Videos](https://varg.ai/how-to/skills/replicate-prompt-videos.md): Prompt video models like a director: shots, camera motion, four audio layers, dialogue, time-coded beats. - [Video Prompting Guide](https://varg.ai/how-to/skills/inference-sh-video-prompting-guide.md): Write prompts that video models follow: shot types, camera moves and per-model tips from Veo to Sora. - [FLUX Image Best Practices](https://varg.ai/how-to/skills/flux-image-best-practices.md): Prompt FLUX image models well: structure, JSON prompts, hex colors, typography and multi-reference edits. - [AI Film Director (Visual Skills)](https://varg.ai/how-to/skills/visual-skills-video-director.md): AI film director for video prompts: dramaturgy first (Murch, blocking, montage), then exact syntax per model. - [FLUX Image Best Practices](https://varg.ai/how-to/skills/flux-image-best-practices.md): Prompt FLUX image models well: structure, JSON prompts, hex colors, typography and multi-reference edits. Prompt FLUX image models well: structure, JSON prompts, hex colors, typography and multi-reference edits. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Prompts by model](https://varg.ai/how-to/browse/skills/prompting.md) - **Slug:** `flux-image-best-practices` - **Status:** live - **Publisher:** Black Forest Labs (vendor, verified) - **License:** MIT - **Source:** https://github.com/black-forest-labs/skills/tree/master/skills/flux-image-best-practices - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** BFL API key (BFL_API_KEY) - **Cost:** Billed by Black Forest Labs (API credits) - **Adoption:** 332 installs, 135 GitHub stars (skills.sh, 2026-10-26) ##### What it does Prompting rules for BFL's FLUX image models across the FLUX.2 and FLUX.1 families, starting from a prompt-structure formula and model selection. Rule files cover text-to-image and image-to-image prompting, structured JSON prompts, hex-color control, typography, multi-reference editing and alternatives to negative prompts. ##### Use when - Writing prompts for FLUX - Exact brand colors or legible text in an image - Editing with several reference images ##### Install - Claude Code: `/plugin marketplace add black-forest-labs/skills && /plugin install flux-image-best-practices@black-forest-labs` - Codex: `npx skills add black-forest-labs/skills --skill flux-image-best-practices -a codex` - Cursor: `npx skills add black-forest-labs/skills --skill flux-image-best-practices -a cursor` - Gemini CLI: `npx skills add black-forest-labs/skills --skill flux-image-best-practices -a gemini-cli` **Models it drives:** flux-2, flux-1 ##### Example output https://s3.varg.ai/how-to/previews/1daa67491a84/flux-image-best-practices.webp — FLUX typography example from BFL's 'Typography & design' prompting guide: a retro hand-lettered poster reading 'If you love me let me sleep' (credit: Black Forest Labs, https://docs.bfl.ai/guides/usecases_t2i_typography_design) ##### More prompts by model skills - [fal Prompting](https://varg.ai/how-to/skills/fal-prompting.md): Prompt specific model families well: Kling, GPT Image 2 and Happy Horse guides, loaded only when needed. - [Replicate Prompt Images](https://varg.ai/how-to/skills/replicate-prompt-images.md): Write better image prompts and edits: photographic language, lighting, text rendering, consistent characters. - [Replicate Prompt Videos](https://varg.ai/how-to/skills/replicate-prompt-videos.md): Prompt video models like a director: shots, camera motion, four audio layers, dialogue, time-coded beats. - [Video Prompting Guide](https://varg.ai/how-to/skills/inference-sh-video-prompting-guide.md): Write prompts that video models follow: shot types, camera moves and per-model tips from Veo to Sora. - [FLUX 3 Prompt Doctor](https://varg.ai/how-to/skills/flux-3-prompt-doctor.md): Diagnose a FLUX 3 video brief before spending: resolve missing, conflicting or schema-changing asks. - [AI Film Director (Visual Skills)](https://varg.ai/how-to/skills/visual-skills-video-director.md): AI film director for video prompts: dramaturgy first (Murch, blocking, montage), then exact syntax per model. - [AI Film Director (Visual Skills)](https://varg.ai/how-to/skills/visual-skills-video-director.md): AI film director for video prompts: dramaturgy first (Murch, blocking, montage), then exact syntax per model. AI film director for video prompts: dramaturgy first (Murch, blocking, montage), then exact syntax per model. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Prompts by model](https://varg.ai/how-to/browse/skills/prompting.md) - **Slug:** `visual-skills-video-director` - **Status:** live - **Publisher:** Serge Shima (community) - **License:** CC-BY-4.0 - **Source:** https://github.com/smixs/visual-skills/tree/main/video - **Works in:** Claude Code, Codex, Cursor - **Cost:** Free — prompts only; generation billed by your video platform - **Adoption:** 971 installs, 582 GitHub stars (GitHub + skills.sh, 2026-10-26) ##### What it does Serge Shima's video skill makes the agent write prompts like a director. It reads a dramaturgy file (Walter Murch's editing priorities, blocking, dynamic montage) and universal rules first, then exactly one model file (Seedance 2.5, Kling 3.0 Turbo/Omni, Veo 3.1 and others), then only the task references that apply (animatic keyframes, camera and lighting vocabulary, race and speed, genre patterns, fixes and skeletons). It covers storyboards, shot lists, director treatments, multi-clip story structure, character continuity, dialogue and sound design, and ends with a dramaturgy check and a three-detail check. A sibling image skill handles Nano Banana and GPT Image 2.5 prompts. ##### Use when - Write, improve, audit or split prompts for Seedance, Kling, Veo, Runway, Luma, Pika or Sora - Turn a script into a storyboard and shot-by-shot prompts - Fix a video prompt that keeps failing ##### Not for - Still images (use the sibling image skill) - Editing or grading footage that already exists ##### Install - Claude Code: `npx skills add smixs/visual-skills@video` - Codex: `npx skills add smixs/visual-skills -g -a codex` **Models it drives:** seedance_2_5, kling_v3, veo_3_1 ##### More prompts by model skills - [fal Prompting](https://varg.ai/how-to/skills/fal-prompting.md): Prompt specific model families well: Kling, GPT Image 2 and Happy Horse guides, loaded only when needed. - [Replicate Prompt Images](https://varg.ai/how-to/skills/replicate-prompt-images.md): Write better image prompts and edits: photographic language, lighting, text rendering, consistent characters. - [Replicate Prompt Videos](https://varg.ai/how-to/skills/replicate-prompt-videos.md): Prompt video models like a director: shots, camera motion, four audio layers, dialogue, time-coded beats. - [Video Prompting Guide](https://varg.ai/how-to/skills/inference-sh-video-prompting-guide.md): Write prompts that video models follow: shot types, camera moves and per-model tips from Veo to Sora. - [FLUX 3 Prompt Doctor](https://varg.ai/how-to/skills/flux-3-prompt-doctor.md): Diagnose a FLUX 3 video brief before spending: resolve missing, conflicting or schema-changing asks. - [FLUX Image Best Practices](https://varg.ai/how-to/skills/flux-image-best-practices.md): Prompt FLUX image models well: structure, JSON prompts, hex colors, typography and multi-reference edits. - [Higgsfield Prompt Skill (O-Side Media)](https://varg.ai/how-to/skills/higgsfield-prompt-skill-oside.md): Independent 36-skill prompt pack for Higgsfield: model picks, camera presets, Soul ID, Cinema Studio, Seedance 2.5. Independent 36-skill prompt pack for Higgsfield: model picks, camera presets, Soul ID, Cinema Studio, Seedance 2.5. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Prompts by model](https://varg.ai/how-to/browse/skills/prompting.md) - **Slug:** `higgsfield-prompt-skill-oside` - **Status:** live - **Publisher:** O-Side Media (community) - **License:** MIT - **Source:** https://github.com/OSideMedia/higgsfield-ai-prompt-skill - **Works in:** Claude Code, Claude.ai, Cursor, Codex - **Requires:** Higgsfield account (generation and credits), Optional: Python 3 for the lint and preflight scripts - **Cost:** Prompts only; generation is billed in Higgsfield credits - **Adoption:** 799 GitHub stars (GitHub, 2026-10-26) ##### What it does An independent prompt pack for Higgsfield users, not made by Higgsfield. A root skill writes and refines video and image prompts with its MCSLA formula, picks the model, camera controls and motion presets for each shot, and routes to 35 sub-skills covering Seedance 2.5 and 2.0 (omni-reference, video edit and extend, VFX), Cinema Studio 2.5/3.0/3.5, Kling 3.0 Motion Control, Soul ID character consistency, an acting system, a shot-list director, a feature-film pipeline, 3D, audio, moodboards and troubleshooting. It keeps a generation ledger, lints claims and Seedance prompts with Python scripts, and ships updates almost weekly (v3.42.0 on 2026-10-23). ##### Use when - Write or fix prompts for any Higgsfield model - Pick camera controls and motion presets per shot - Keep a character consistent with Soul ID ##### Not for - Generating without a Higgsfield account - Other platforms (prompts assume Higgsfield's model menu) ##### Install - Claude Code: `git clone https://github.com/OSideMedia/higgsfield-ai-prompt-skill ~/.claude/skills/higgsfield` **Models it drives:** seedance_2_5, seedance_2, kling_v3, veo_3_1, nano_banana_pro, soul ##### More prompts by model skills - [fal Prompting](https://varg.ai/how-to/skills/fal-prompting.md): Prompt specific model families well: Kling, GPT Image 2 and Happy Horse guides, loaded only when needed. - [Replicate Prompt Images](https://varg.ai/how-to/skills/replicate-prompt-images.md): Write better image prompts and edits: photographic language, lighting, text rendering, consistent characters. - [Replicate Prompt Videos](https://varg.ai/how-to/skills/replicate-prompt-videos.md): Prompt video models like a director: shots, camera motion, four audio layers, dialogue, time-coded beats. - [Video Prompting Guide](https://varg.ai/how-to/skills/inference-sh-video-prompting-guide.md): Write prompts that video models follow: shot types, camera moves and per-model tips from Veo to Sora. - [FLUX 3 Prompt Doctor](https://varg.ai/how-to/skills/flux-3-prompt-doctor.md): Diagnose a FLUX 3 video brief before spending: resolve missing, conflicting or schema-changing asks. - [FLUX Image Best Practices](https://varg.ai/how-to/skills/flux-image-best-practices.md): Prompt FLUX image models well: structure, JSON prompts, hex colors, typography and multi-reference edits. - [Seedance 2.0 Prompt Guide](https://varg.ai/how-to/skills/seedance-2-prompt-guide.md): Prompt guide for Seedance 2.0 multimodal video: the @ reference system, camera language and ready templates. Prompt guide for Seedance 2.0 multimodal video: the @ reference system, camera language and ready templates. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Prompts by model](https://varg.ai/how-to/browse/skills/prompting.md) - **Slug:** `seedance-2-prompt-guide` - **Status:** live - **Publisher:** Dex Hünter (community) - **License:** MIT - **Source:** https://github.com/dexhunter/seedance2-skill - **Works in:** Claude Code, Codex, Cursor - **Cost:** Free — prompts only; generation billed by your Seedance provider - **Adoption:** 3,124 installs, 4,363 GitHub stars (GitHub + skills.sh, 2026-10-26) ##### What it does Teaches the agent to write prompts for ByteDance's Seedance 2.0 (Jimeng/Dreamina) from text, image, video and audio inputs: input limits, the @ reference system and role assignment, a prompt formula with time-segmented prompts for clips over 10 seconds, a camera-language reference, and patterns for character consistency, camera-move replication, effect templates, video extension and editing, music beat-matching, dialogue, one-take shots, e-commerce and educational content. Ships in English and Chinese (seedance-prompt-zh) and is one of the most-starred Seedance skills (4.4K stars). ##### Use when - Write a Seedance 2.0 prompt from image, clip or audio references - Replicate a camera move or effect from a reference video - Extend or edit an existing Seedance clip ##### Not for - Seedance 2.5-specific features (written for 2.0) ##### Install - Claude Code: `npx skills add dexhunter/seedance2-skill` **Models it drives:** seedance_2 ##### How-to tutorials that use it - [How to make exercise demo video loops with AI](https://varg.ai/how-to/make-exercise-demo-video-loops.md): Looping exercise demos for a fitness app: one trainer, many exercises, clean background. - [How to remake a reference video with AI](https://varg.ai/how-to/remake-a-reference-video.md): Take a reel or ad you like, break it down shot by shot, and remake the format for your brand. - [How to swap your character into a video with AI](https://varg.ai/how-to/swap-your-character-into-a-video.md): Your character (or person) performs the exact motion of a trending video: dance, gesture, walk, ad performance — motion transfer. ##### More prompts by model skills - [fal Prompting](https://varg.ai/how-to/skills/fal-prompting.md): Prompt specific model families well: Kling, GPT Image 2 and Happy Horse guides, loaded only when needed. - [Replicate Prompt Images](https://varg.ai/how-to/skills/replicate-prompt-images.md): Write better image prompts and edits: photographic language, lighting, text rendering, consistent characters. - [Replicate Prompt Videos](https://varg.ai/how-to/skills/replicate-prompt-videos.md): Prompt video models like a director: shots, camera motion, four audio layers, dialogue, time-coded beats. - [Video Prompting Guide](https://varg.ai/how-to/skills/inference-sh-video-prompting-guide.md): Write prompts that video models follow: shot types, camera moves and per-model tips from Veo to Sora. - [FLUX 3 Prompt Doctor](https://varg.ai/how-to/skills/flux-3-prompt-doctor.md): Diagnose a FLUX 3 video brief before spending: resolve missing, conflicting or schema-changing asks. - [FLUX Image Best Practices](https://varg.ai/how-to/skills/flux-image-best-practices.md): Prompt FLUX image models well: structure, JSON prompts, hex colors, typography and multi-reference edits. - [Awesome Seedance Skills](https://varg.ai/how-to/skills/awesome-seedance-skills.md): Evidence-led Seedance 2.5/2.0 skills: a production workflow plus genre templates distilled from 934 verified cases. Evidence-led Seedance 2.5/2.0 skills: a production workflow plus genre templates distilled from 934 verified cases. - **Type:** [Prompt](https://varg.ai/how-to/prompts.md) · [Prompts by model](https://varg.ai/how-to/browse/skills/prompting.md) - **Slug:** `awesome-seedance-skills` - **Status:** live - **Publisher:** LearnPrompt (community) - **License:** MIT - **Source:** https://github.com/LearnPrompt/awesome-seedance/tree/main/agents/skills - **Works in:** Claude Code, Codex - **Cost:** Free — prompts and plans; generation billed by your Seedance provider - **Adoption:** 2,294 GitHub stars (GitHub, 2026-10-26) ##### What it does From the awesome-seedance library (934 cases checked against their original posts, 495 cross-model retests, 27 templates, synced daily from goodcase.ai). The repo's own skills include a production workflow that carries a Seedance video from brief through references, storyboard, generation handoff, shot-level review and targeted repair (with a plan validator), a prompt-library skill, a storyboard-grid-to-video skill, and genre template skills (3D cartoon, vehicles, epic fantasy and sci-fi, fashion lookbook, food ASMR, horror, memes, pets, retro DV home video, sports, travel walks). Template skills write a clip as locked identities plus a timecoded cut list. ##### Use when - Plan and review a multi-shot Seedance video end to end - Write a genre-specific Seedance prompt from a proven structure ##### Not for - Models other than Seedance (structures were tested on Seedance 2.5 and 2.0) ##### Install - Claude Code: `npx skills add LearnPrompt/awesome-seedance --skill seedance-production-workflow` - Codex: `npx skills add LearnPrompt/awesome-seedance --skill seedance-production-workflow` **Models it drives:** seedance_2_5, seedance_2 ##### How-to tutorials that use it - [How to make an AI short film](https://varg.ai/how-to/make-an-ai-short-film.md): A multi-shot story with consistent characters: script, storyboard, previz, shots, edit. - [How to keep a character consistent in AI video](https://varg.ai/how-to/keep-a-character-consistent-in-ai-video.md): One mascot, trainer or actor that looks identical across many images and clips. - [How to make trending AI video formats](https://varg.ai/how-to/make-trending-ai-video-formats.md): What's going viral now — formats you can make for your brand today, with the skill for each. ##### More prompts by model skills - [fal Prompting](https://varg.ai/how-to/skills/fal-prompting.md): Prompt specific model families well: Kling, GPT Image 2 and Happy Horse guides, loaded only when needed. - [Replicate Prompt Images](https://varg.ai/how-to/skills/replicate-prompt-images.md): Write better image prompts and edits: photographic language, lighting, text rendering, consistent characters. - [Replicate Prompt Videos](https://varg.ai/how-to/skills/replicate-prompt-videos.md): Prompt video models like a director: shots, camera motion, four audio layers, dialogue, time-coded beats. - [Video Prompting Guide](https://varg.ai/how-to/skills/inference-sh-video-prompting-guide.md): Write prompts that video models follow: shot types, camera moves and per-model tips from Veo to Sora. - [FLUX 3 Prompt Doctor](https://varg.ai/how-to/skills/flux-3-prompt-doctor.md): Diagnose a FLUX 3 video brief before spending: resolve missing, conflicting or schema-changing asks. - [FLUX Image Best Practices](https://varg.ai/how-to/skills/flux-image-best-practices.md): Prompt FLUX image models well: structure, JSON prompts, hex colors, typography and multi-reference edits. ## Effects — AI video effects: titles, captions, transitions ### Cinematic & short films - [fal Fan Cam](https://varg.ai/how-to/skills/fal-fan-cam.md): Put someone into a live sports broadcast: a realistic fan-cam cutaway with scoreboard and channel overlays. Put someone into a live sports broadcast: a realistic fan-cam cutaway with scoreboard and channel overlays. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Cinematic & short films](https://varg.ai/how-to/browse/skills/cinematic.md) - **Slug:** `fal-fan-cam` - **Status:** live - **Publisher:** fal (vendor, verified) - **License:** MIT - **Source:** https://github.com/fal-ai-community/skills/tree/main/skills/fan-cam - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** fal API key (FAL_KEY), genmedia CLI - **Cost:** Billed by fal per endpoint call - **Adoption:** 271 installs, 256 GitHub stars (skills.sh, 2026-10-26) ##### What it does Creates personalized fan-cam videos from a user photo: a realistic stadium or arena crowd shot made with GPT Image 2, animated with Kling v3, with scoreboard overlays, TV channel bugs and broadcast framing. It includes a prompt contract and a quality gate for identity and realism. ##### Use when - A fan-cam or crowd-reaction clip from a selfie - Broadcast-style sports cutaways ##### Install - Claude Code: `npx skills add fal-ai-community/skills --skill fan-cam -a claude-code` - Codex: `npx skills add fal-ai-community/skills --skill fan-cam -a codex` - Cursor: `npx skills add fal-ai-community/skills --skill fan-cam -a cursor` - Gemini CLI: `npx skills add fal-ai-community/skills --skill fan-cam -a gemini-cli` **Models it drives:** gpt_image_2, kling_v3 ##### More cinematic & short films skills - [Previz to Seedance](https://varg.ai/how-to/skills/varg-previz.md): Plan multi-shot AI video before paying: story, anchor sheets, a three.js grey-box previz, then a Seedance v2v pack. - [Higgsfield Generate](https://varg.ai/how-to/skills/higgsfield-generate.md): Generate images, video, 3D and audio across 30+ models, plus ad videos and virality scoring, from one CLI. - [fal Recipes](https://varg.ai/how-to/skills/fal-recipes.md): Follow proven multi-step fal pipelines for commercials, product shots, characters, lip-sync and try-on. - [Gemini Omni Flash](https://varg.ai/how-to/skills/gemini-omni-flash-api.md): Generate and edit video with Gemini Omni 1.1 Flash: text-to-video, first/last frames, extensions, edits. - [Melies](https://varg.ai/how-to/skills/melies.md): Generate images, video, movie posters and thumbnails from the terminal with 148 AI actors and style flags. - [AI Video Generation](https://varg.ai/how-to/skills/inference-sh-ai-video-generation.md): Generate video with 40+ models from one CLI: Veo 3.1, Seedance 2.0, HappyHorse, Wan, Grok and more. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt ### Animation & motion graphics - [p5 Paint Animation](https://varg.ai/how-to/skills/hyperframes-p5-paint-animation.md): Handwriting that writes itself, photos that paint themselves and clips repainted as living paintings, in p5.js code. Handwriting that writes itself, photos that paint themselves and clips repainted as living paintings, in p5.js code. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Animation & motion graphics](https://varg.ai/how-to/browse/skills/animation.md) - **Slug:** `hyperframes-p5-paint-animation` - **Status:** live - **Publisher:** HeyGen (vendor, verified) - **License:** Apache-2.0 - **Source:** https://github.com/heygen-com/hyperframes-community-skills/tree/master/skills/p5-paint-animation - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** Node.js 22.12+, FFmpeg, About 500 MB of disk for Puppeteer's Chrome for Testing - **Cost:** Free — open-source, renders locally, no image models - **Adoption:** 318 installs, 231 GitHub stars (skills.sh, 2026-10-26) ##### What it does Three flows on one headless p5.brush engine: a phrase writes itself in pencil on ruled paper (with a pen track for camera moves), a photo repaints itself as brushstrokes from coarse to fine, and a short clip or Live Photo is repainted frame by frame over a locked stroke lattice so the painting holds still while the subject moves. Everything is seeded code, so the same seed gives the same frames; sketches run in sandboxed headless Chromium with network blocked and ffmpeg stitches the MP4. Render time grows per frame, so keep footage under about 10 s; a community-tier skill added by HeyGen's James Russo. ##### Use when - Handwritten text that writes itself - Turn a photo into a painting that paints itself - Repaint a short clip or Live Photo as a living painting - Sketch or write-on effects with no image model ##### Not for - Crisp or brand-accurate typography (use real fonts) - Photoreal output - Footage longer than about 10 s - Live or interactive pages ##### Install - Claude Code: `npx skills add heygen-com/hyperframes-community-skills --skill p5-paint-animation -a claude-code` - Codex: `npx skills add heygen-com/hyperframes-community-skills --skill p5-paint-animation -a codex` - Cursor: `npx skills add heygen-com/hyperframes-community-skills --skill p5-paint-animation -a cursor` - Gemini CLI: `npx skills add heygen-com/hyperframes-community-skills --skill p5-paint-animation -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/6e64b6b03bf7/hyperframes-p5-paint-animation.mp4 — From the official community-skills launch film made with the p5 skill: 'p5.js' writes itself by hand, panels paint on, and a beach clip is repainted as a living painting (65-75 s) (credit: HyperFrames (@HyperFrames_), https://x.com/HyperFrames_/status/2092308986775343288) ##### More animation & motion graphics skills - [Day in My Life](https://varg.ai/how-to/skills/hyperframes-day-in-my-life.md): A 60–75 s hand-drawn ink film of your agent's day with you, built from local Claude Code history, scored for strings. - [Session Story](https://varg.ai/how-to/skills/hyperframes-session-story.md): A 40–70 s watercolor film of a typical session with your agent, told in your real messages, with its own score. - [Remotion Markup](https://varg.ai/how-to/skills/remotion-markup.md): Write Remotion markup that animates well: timing, transitions, motion blur, effects, audio, fonts and 3D. - [Motion AI Kit](https://varg.ai/how-to/skills/motion-ai-kit.md): Write production-grade UI animation with Motion or CSS: best practices, docs search, springs, perf audits. - [Motion Design](https://varg.ai/how-to/skills/lottiefiles-motion-design.md): Animate with intent: timing, easing, choreography and Disney principles for UI and motion graphics. - [dotLottie Web](https://varg.ai/how-to/skills/dotlottie-web.md): Ship Lottie animations on the web with dotLottie: package choice, workers, state machines, theming, slots. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Bang Motion](https://varg.ai/how-to/skills/bang-motion.md): Browser motion graphics that move like video, not slides: openers, promos, bumpers, kinetic type and explainers. Browser motion graphics that move like video, not slides: openers, promos, bumpers, kinetic type and explainers. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Animation & motion graphics](https://varg.ai/how-to/browse/skills/animation.md) - **Slug:** `bang-motion` - **Status:** live - **Publisher:** Bang Tutorial (community) - **License:** MIT - **Source:** https://github.com/bangtutorial/bang-motion - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** A browser for preview, Node.js with Puppeteer and ffmpeg for MP4 export - **Cost:** Free — local render - **Adoption:** 149 installs, 733 GitHub stars (GitHub + skills.sh, 2026-10-26) ##### What it does An Indonesian-language skill from the Bang Tutorial YouTube channel for building cinematic motion graphics in a single HTML file with HTML, CSS and GSAP (plus Three.js when needed): promo videos, openers and intros, bumpers, kinetic typography and five explainer styles (visual journalism, cartoon, catalogue, stage and sketch) in 16:9 or 9:16. Its rules are mechanical anti-slideshow checks: deterministic timing, a style derived from the theme rather than the starter, live text over a quiet background. Puppeteer scripts export frames to MP4, and a reference covers building explainers directly in After Effects through Higgsfield's bridge. ##### Use when - Make a promo, opener, bumper or kinetic-type piece in the browser - Fix web animation that looks like a slide deck - Render a web animation to MP4 ##### Not for - AI-generated footage - Teams that need English instructions (SKILL.md is in Indonesian) ##### Install - Claude Code: `/plugin marketplace add bangtutorial/bang-motion && /plugin install bang-motion@bang-motion` ##### Example output https://s3.varg.ai/how-to/previews/1f3a31890598/bang-motion.webp — Frame from the README style gallery: 'Poster color-block' kinetic typography (huge cropped words, color fields that slam in on the beat). Still image; no video demo is published in the repo. (credit: Bang Tutorial (bang-motion), https://github.com/bangtutorial/bang-motion#style-gallery) ##### More animation & motion graphics skills - [Day in My Life](https://varg.ai/how-to/skills/hyperframes-day-in-my-life.md): A 60–75 s hand-drawn ink film of your agent's day with you, built from local Claude Code history, scored for strings. - [p5 Paint Animation](https://varg.ai/how-to/skills/hyperframes-p5-paint-animation.md): Handwriting that writes itself, photos that paint themselves and clips repainted as living paintings, in p5.js code. - [Session Story](https://varg.ai/how-to/skills/hyperframes-session-story.md): A 40–70 s watercolor film of a typical session with your agent, told in your real messages, with its own score. - [Remotion Markup](https://varg.ai/how-to/skills/remotion-markup.md): Write Remotion markup that animates well: timing, transitions, motion blur, effects, audio, fonts and 3D. - [Motion AI Kit](https://varg.ai/how-to/skills/motion-ai-kit.md): Write production-grade UI animation with Motion or CSS: best practices, docs search, springs, perf audits. - [Motion Design](https://varg.ai/how-to/skills/lottiefiles-motion-design.md): Animate with intent: timing, easing, choreography and Disney principles for UI and motion graphics. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt ### Launch videos - [Orbit Card](https://varg.ai/how-to/skills/hyperframes-orbit-card.md): A single feature card orbits a dot sphere on a Blender-made camera move: a 10-second product beat. A single feature card orbits a dot sphere on a Blender-made camera move: a 10-second product beat. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Launch videos](https://varg.ai/how-to/browse/skills/launch.md) - **Slug:** `hyperframes-orbit-card` - **Status:** live - **Publisher:** HeyGen (vendor, verified) - **License:** Apache-2.0 - **Source:** https://github.com/heygen-com/hyperframes/tree/main/registry/blocks/orbit-card - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** HyperFrames project (Node.js 22+, FFmpeg) - **Cost:** Free — open-source, renders locally - **Adoption:** 10,100 installs, 73,064 GitHub stars (skills.sh, 2026-10-26) ##### What it does A HyperFrames registry block: a single feature card orbits a dot sphere on approved Blender camera motion (1920×1080, 10 s, 4 variables). Install it into a project with `npx hyperframes add orbit-card`, mount it from `index.html` and set its variables; it renders locally like any other composition. ##### Use when - A ready-made title or motion moment instead of hand-built animation - Openers, chapter cards and launch-film beats in a HyperFrames project ##### Install - Claude Code: `npx skills add heygen-com/hyperframes --skill orbit-card -a claude-code` - Codex: `npx skills add heygen-com/hyperframes --skill orbit-card -a codex` - Cursor: `npx skills add heygen-com/hyperframes --skill orbit-card -a cursor` - Gemini CLI: `npx skills add heygen-com/hyperframes --skill orbit-card -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/d22e16d165d1/hyperframes-orbit-card.mp4 — Official catalog render of the block: the 'Always in sync' feature card with a dot sphere orbits on the camera move (credit: HeyGen (HyperFrames), https://hyperframes.heygen.com/catalog/blocks/orbit-card) ##### More launch videos skills - [Onetake](https://varg.ai/how-to/skills/onetake.md): Launch films that never cut to the next slide: one continuous camera, measured moves, motion blur, a slideshow oracle. - [Guizang Product Video](https://varg.ai/how-to/skills/guizang-product-video.md): Software promos and release-notes videos built from your real product components, with original music and SFX. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt ### Explainers - [Remotion Maps](https://varg.ai/how-to/skills/remotion-maps.md): Animate maps in Remotion: static maps, Mapbox, MapLibre, MapTiler and CesiumJS 3D flyovers. Animate maps in Remotion: static maps, Mapbox, MapLibre, MapTiler and CesiumJS 3D flyovers. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Explainers](https://varg.ai/how-to/browse/skills/explainers.md) - **Slug:** `remotion-maps` - **Status:** live - **Publisher:** Remotion (vendor, verified) - **License:** MIT - **Source:** https://github.com/remotion-dev/skills/tree/main/skills/remotion-maps - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** Node.js, Remotion 4, Mapbox, MapTiler or Cesium token for those techniques - **Cost:** Free — local render (Remotion company license for larger for-profit teams) - **Adoption:** 189,700 installs, 5,750 GitHub stars (skills.sh, 2026-10-26) ##### What it does A technique library for map animation in Remotion with one folder per approach: static maps, Mapbox, MapLibre, MapTiler and CesiumJS. Each technique ships a TECHNIQUE.md, TSX components, sample GeoJSON and path data, and prep scripts for routes and 3D flyovers. ##### Use when - Route, travel or location animations - Data stories on a map - 3D city or terrain flyovers ##### Install - Claude Code: `npx skills add remotion-dev/skills --skill remotion-maps -a claude-code` - Codex: `npx skills add remotion-dev/skills --skill remotion-maps -a codex` - Cursor: `npx skills add remotion-dev/skills --skill remotion-maps -a cursor` - Gemini CLI: `npx skills add remotion-dev/skills --skill remotion-maps -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/e38ee7b9062f/remotion-maps.mp4 — Maps skill launch demo: the prompt, then the map zooms out of Los Angeles and a route animates to New York with the camera following it (credit: Jonny Burger (@JNYBGR), Remotion founder, https://x.com/JNYBGR/status/2015059120080977936) ##### How-to tutorials that use it - [How to make an animated chart video with AI](https://varg.ai/how-to/make-an-animated-chart-video.md): Numbers from a CSV or a report as an animated chart: bar race, line, hero number. ##### More explainers skills - [HyperFrames](https://varg.ai/how-to/skills/hyperframes.md): Turn HTML, CSS and seekable animation into deterministic MP4s: launch films, explainers, captions, title cards. - [HyperFrames Slideshow](https://varg.ai/how-to/skills/hyperframes-slideshow.md): Build a live, navigable deck with fragment reveals, branching hotspots and presenter notes instead of an MP4. - [Plain-Sight Explainer](https://varg.ai/how-to/skills/hyperframes-vox-explainer.md): 60–90 s collage explainers about everyday things, or mined from your documents and links, with numeric seam checks. - [Remotion Create](https://varg.ai/how-to/skills/remotion-create.md): Scaffold a new Remotion project and lay the video out as clips or scenes before writing any frames. - [Higgsfield Video Explainer](https://varg.ai/how-to/skills/higgsfield-video-explainer.md): Turn a topic or document into a narrated, stylized explainer assembled from 10-second voiced blocks. - [Explainer Video Guide](https://varg.ai/how-to/skills/inference-sh-explainer-video-guide.md): Plan explainer videos end to end: script formulas, pacing, scene planning, voiceover and assembly. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt ### Text & titles - [Canopy Part Title](https://varg.ai/how-to/skills/hyperframes-canopy-part-title.md): Leaves sweep across the frame and part to reveal your headline: a 12-second 1080p title block. Leaves sweep across the frame and part to reveal your headline: a 12-second 1080p title block. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Text & titles](https://varg.ai/how-to/browse/skills/kinetic-type.md) - **Slug:** `hyperframes-canopy-part-title` - **Status:** live - **Publisher:** HeyGen (vendor, verified) - **License:** Apache-2.0 - **Source:** https://github.com/heygen-com/hyperframes/tree/main/registry/blocks/canopy-part-title - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** HyperFrames project (Node.js 22+, FFmpeg) - **Cost:** Free — open-source, renders locally - **Adoption:** 10,100 installs, 73,064 GitHub stars (skills.sh, 2026-10-26) ##### What it does A HyperFrames registry block: leaves sweep through the frame and part to reveal the headline (1920×1080, 12 s, 11 variables). Install it into a project with `npx hyperframes add canopy-part-title`, mount it from `index.html` and set its variables; it renders locally like any other composition. ##### Use when - A ready-made title or motion moment instead of hand-built animation - Openers, chapter cards and launch-film beats in a HyperFrames project ##### Install - Claude Code: `npx skills add heygen-com/hyperframes --skill canopy-part-title -a claude-code` - Codex: `npx skills add heygen-com/hyperframes --skill canopy-part-title -a codex` - Cursor: `npx skills add heygen-com/hyperframes --skill canopy-part-title -a cursor` - Gemini CLI: `npx skills add heygen-com/hyperframes --skill canopy-part-title -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/28caaca90886/hyperframes-canopy-part-title.mp4 — Official catalog render of the block: leaves sweep in and part to reveal 'Understory', then a second sweep swaps to 'Move slowly' (credit: HeyGen (HyperFrames), https://hyperframes.heygen.com/catalog/blocks/canopy-part-title) ##### More text & titles skills - [Code Slice Hero](https://varg.ai/how-to/skills/hyperframes-code-slice-hero.md): A tiled headline flips cell by cell under a sweeping depth field to reveal a second line: an 8-second hero. - [Glass Shard Title](https://varg.ai/how-to/skills/hyperframes-glass-shard-title.md): Glass shards fly in through fog and tile themselves into your headline, then fly out past the camera. - [Wireframe Portal Title](https://varg.ai/how-to/skills/hyperframes-wireframe-portal-title.md): A wireframe portal bursts open, the title comes through, then its letters swap into a second phrase. - [PixiJS Text](https://varg.ai/how-to/skills/pixijs-scene-text.md): Render and animate text in PixiJS v8: canvas Text, BitmapText, HTMLText and per-character SplitText. - [Kinetic Typography](https://varg.ai/how-to/skills/iart-kinetic-typography.md): Kinetic type for web and video: line, word or char splits, mask and blur reveals, variable-font motion, frozen-frame QA. - [Title & Kinetic Typography Direction](https://varg.ai/how-to/skills/calesthio-title-kinetic-typography.md): Typography direction for video: text jobs, hierarchy, reveal-hold-exit timing, safe zones, contrast, font licensing. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Code Slice Hero](https://varg.ai/how-to/skills/hyperframes-code-slice-hero.md): A tiled headline flips cell by cell under a sweeping depth field to reveal a second line: an 8-second hero. A tiled headline flips cell by cell under a sweeping depth field to reveal a second line: an 8-second hero. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Text & titles](https://varg.ai/how-to/browse/skills/kinetic-type.md) - **Slug:** `hyperframes-code-slice-hero` - **Status:** live - **Publisher:** HeyGen (vendor, verified) - **License:** Apache-2.0 - **Source:** https://github.com/heygen-com/hyperframes/tree/main/registry/blocks/code-slice-hero - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** HyperFrames project (Node.js 22+, FFmpeg) - **Cost:** Free — open-source, renders locally - **Adoption:** 10,100 installs, 73,064 GitHub stars (skills.sh, 2026-10-26) ##### What it does A HyperFrames registry block: a tiled headline surface flips cell by cell under a sweeping depth field to reveal the rear headline (1920×1080, 8 s, 18 variables). Install it into a project with `npx hyperframes add code-slice-hero`, mount it from `index.html` and set its variables; it renders locally like any other composition. ##### Use when - A ready-made title or motion moment instead of hand-built animation - Openers, chapter cards and launch-film beats in a HyperFrames project ##### Install - Claude Code: `npx skills add heygen-com/hyperframes --skill code-slice-hero -a claude-code` - Codex: `npx skills add heygen-com/hyperframes --skill code-slice-hero -a codex` - Cursor: `npx skills add heygen-com/hyperframes --skill code-slice-hero -a cursor` - Gemini CLI: `npx skills add heygen-com/hyperframes --skill code-slice-hero -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/9f799bc52087/hyperframes-code-slice-hero.mp4 — Official catalog render of the block: the tiled 'MAKE IT' headline flips cell by cell to reveal 'MATTER.' (credit: HeyGen (HyperFrames), https://hyperframes.heygen.com/catalog/blocks/code-slice-hero) ##### More text & titles skills - [Canopy Part Title](https://varg.ai/how-to/skills/hyperframes-canopy-part-title.md): Leaves sweep across the frame and part to reveal your headline: a 12-second 1080p title block. - [Glass Shard Title](https://varg.ai/how-to/skills/hyperframes-glass-shard-title.md): Glass shards fly in through fog and tile themselves into your headline, then fly out past the camera. - [Wireframe Portal Title](https://varg.ai/how-to/skills/hyperframes-wireframe-portal-title.md): A wireframe portal bursts open, the title comes through, then its letters swap into a second phrase. - [PixiJS Text](https://varg.ai/how-to/skills/pixijs-scene-text.md): Render and animate text in PixiJS v8: canvas Text, BitmapText, HTMLText and per-character SplitText. - [Kinetic Typography](https://varg.ai/how-to/skills/iart-kinetic-typography.md): Kinetic type for web and video: line, word or char splits, mask and blur reveals, variable-font motion, frozen-frame QA. - [Title & Kinetic Typography Direction](https://varg.ai/how-to/skills/calesthio-title-kinetic-typography.md): Typography direction for video: text jobs, hierarchy, reveal-hold-exit timing, safe zones, contrast, font licensing. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Glass Shard Title](https://varg.ai/how-to/skills/hyperframes-glass-shard-title.md): Glass shards fly in through fog and tile themselves into your headline, then fly out past the camera. Glass shards fly in through fog and tile themselves into your headline, then fly out past the camera. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Text & titles](https://varg.ai/how-to/browse/skills/kinetic-type.md) - **Slug:** `hyperframes-glass-shard-title` - **Status:** live - **Publisher:** HeyGen (vendor, verified) - **License:** Apache-2.0 - **Source:** https://github.com/heygen-com/hyperframes/tree/main/registry/blocks/glass-shard-title - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** HyperFrames project (Node.js 22+, FFmpeg) - **Cost:** Free — open-source, renders locally - **Adoption:** 10,100 installs, 73,064 GitHub stars (skills.sh, 2026-10-26) ##### What it does A HyperFrames registry block: glass shards fly in through fog and tile themselves into the headline (1920×1080, 12.16 s, 19 variables). Install it into a project with `npx hyperframes add glass-shard-title`, mount it from `index.html` and set its variables; it renders locally like any other composition. ##### Use when - A ready-made title or motion moment instead of hand-built animation - Openers, chapter cards and launch-film beats in a HyperFrames project ##### Install - Claude Code: `npx skills add heygen-com/hyperframes --skill glass-shard-title -a claude-code` - Codex: `npx skills add heygen-com/hyperframes --skill glass-shard-title -a codex` - Cursor: `npx skills add heygen-com/hyperframes --skill glass-shard-title -a cursor` - Gemini CLI: `npx skills add heygen-com/hyperframes --skill glass-shard-title -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/f6b92a1beeaa/hyperframes-glass-shard-title.mp4 — Official catalog render of the block: glass shards fly in, tile into 'Designed in glass', then fly out past the camera (credit: HeyGen (HyperFrames), https://hyperframes.heygen.com/catalog/blocks/glass-shard-title) ##### More text & titles skills - [Canopy Part Title](https://varg.ai/how-to/skills/hyperframes-canopy-part-title.md): Leaves sweep across the frame and part to reveal your headline: a 12-second 1080p title block. - [Code Slice Hero](https://varg.ai/how-to/skills/hyperframes-code-slice-hero.md): A tiled headline flips cell by cell under a sweeping depth field to reveal a second line: an 8-second hero. - [Wireframe Portal Title](https://varg.ai/how-to/skills/hyperframes-wireframe-portal-title.md): A wireframe portal bursts open, the title comes through, then its letters swap into a second phrase. - [PixiJS Text](https://varg.ai/how-to/skills/pixijs-scene-text.md): Render and animate text in PixiJS v8: canvas Text, BitmapText, HTMLText and per-character SplitText. - [Kinetic Typography](https://varg.ai/how-to/skills/iart-kinetic-typography.md): Kinetic type for web and video: line, word or char splits, mask and blur reveals, variable-font motion, frozen-frame QA. - [Title & Kinetic Typography Direction](https://varg.ai/how-to/skills/calesthio-title-kinetic-typography.md): Typography direction for video: text jobs, hierarchy, reveal-hold-exit timing, safe zones, contrast, font licensing. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Wireframe Portal Title](https://varg.ai/how-to/skills/hyperframes-wireframe-portal-title.md): A wireframe portal bursts open, the title comes through, then its letters swap into a second phrase. A wireframe portal bursts open, the title comes through, then its letters swap into a second phrase. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Text & titles](https://varg.ai/how-to/browse/skills/kinetic-type.md) - **Slug:** `hyperframes-wireframe-portal-title` - **Status:** live - **Publisher:** HeyGen (vendor, verified) - **License:** Apache-2.0 - **Source:** https://github.com/heygen-com/hyperframes/tree/main/registry/blocks/wireframe-portal-title - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** HyperFrames project (Node.js 22+, FFmpeg) - **Cost:** Free — open-source, renders locally - **Adoption:** 10,100 installs, 73,064 GitHub stars (skills.sh, 2026-10-26) ##### What it does A HyperFrames registry block: a wireframe portal bursts open, the title comes through, then its letters swap into a second phrase (1920×1080, 8 s, 9 variables). Install it into a project with `npx hyperframes add wireframe-portal-title`, mount it from `index.html` and set its variables; it renders locally like any other composition. ##### Use when - A ready-made title or motion moment instead of hand-built animation - Openers, chapter cards and launch-film beats in a HyperFrames project ##### Install - Claude Code: `npx skills add heygen-com/hyperframes --skill wireframe-portal-title -a claude-code` - Codex: `npx skills add heygen-com/hyperframes --skill wireframe-portal-title -a codex` - Cursor: `npx skills add heygen-com/hyperframes --skill wireframe-portal-title -a cursor` - Gemini CLI: `npx skills add heygen-com/hyperframes --skill wireframe-portal-title -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/a9084c5eed14/hyperframes-wireframe-portal-title.mp4 — Official catalog render of the block: 'BREAKTHROUGH' bursts through a wireframe portal, then its letters swap to a second phrase (credit: HeyGen (HyperFrames), https://hyperframes.heygen.com/catalog/blocks/wireframe-portal-title) ##### More text & titles skills - [Canopy Part Title](https://varg.ai/how-to/skills/hyperframes-canopy-part-title.md): Leaves sweep across the frame and part to reveal your headline: a 12-second 1080p title block. - [Code Slice Hero](https://varg.ai/how-to/skills/hyperframes-code-slice-hero.md): A tiled headline flips cell by cell under a sweeping depth field to reveal a second line: an 8-second hero. - [Glass Shard Title](https://varg.ai/how-to/skills/hyperframes-glass-shard-title.md): Glass shards fly in through fog and tile themselves into your headline, then fly out past the camera. - [PixiJS Text](https://varg.ai/how-to/skills/pixijs-scene-text.md): Render and animate text in PixiJS v8: canvas Text, BitmapText, HTMLText and per-character SplitText. - [Kinetic Typography](https://varg.ai/how-to/skills/iart-kinetic-typography.md): Kinetic type for web and video: line, word or char splits, mask and blur reveals, variable-font motion, frozen-frame QA. - [Title & Kinetic Typography Direction](https://varg.ai/how-to/skills/calesthio-title-kinetic-typography.md): Typography direction for video: text jobs, hierarchy, reveal-hold-exit timing, safe zones, contrast, font licensing. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Kinetic Typography](https://varg.ai/how-to/skills/iart-kinetic-typography.md): Kinetic type for web and video: line, word or char splits, mask and blur reveals, variable-font motion, frozen-frame QA. Kinetic type for web and video: line, word or char splits, mask and blur reveals, variable-font motion, frozen-frame QA. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Text & titles](https://varg.ai/how-to/browse/skills/kinetic-type.md) - **Slug:** `iart-kinetic-typography` - **Status:** live - **Publisher:** iart.ai (community) - **License:** MIT - **Source:** https://github.com/iart-ai/kinetic-typography-skills/tree/main/skills/kinetic-typography - **Works in:** Claude Code, Cursor, Codex - **Requires:** A Chromium browser for the standalone-HTML tier, or Node.js with Remotion for video, Optional helpers: Playwright via npx, ffmpeg and ffprobe - **Cost:** Free — open-source, renders locally - **Adoption:** 2,313 installs, 16 GitHub stars (skills.sh, 2026-10-26) ##### What it does Animates text for web and video: pick line, word or character granularity, then apply mask, blur, clip-path or stagger reveals with concrete easing and stagger timings, animate variable-font axes through `font-variation-settings`, or run SVG text along a path, with recipes for CSS, GSAP SplitText, Framer Motion and Remotion. For HTML deliverables it requires splitting only after `document.fonts.ready`, a `?t=N` freeze harness for deterministic screenshots, and a start/mid/end check for clipped glyphs, FOUC and wrong line breaks. The verify helpers (Playwright screenshots, an ffmpeg contact sheet, an ffprobe spec check) live in the repo-root `scripts/`, so they arrive with the Claude Code plugin install but not with a skills-CLI install that copies only the skill folder. Per-character splitting breaks Arabic-script joining; use Persian Motion Director for RTL text. ##### Use when - Animate a headline or an animated title card - Split-text reveal or stagger by character, word or line - Lyric or caption video with animated words - Animate a variable font's weight or put text on a path ##### Not for - Arabic-script or RTL text: per-letter boxes break joining (use persian-motion-director) - Transcript-driven subtitle tracks (use remotion-captions or a captions skill) - Choosing or licensing fonts (use adobe-fonts or chinese-font-selector) ##### Install - Claude Code: `/plugin marketplace add iart-ai/kinetic-typography-skills && /plugin install kinetic-typography-skills` - Codex: `npx skills add iart-ai/kinetic-typography-skills` - Cursor: `npx skills add iart-ai/kinetic-typography-skills` ##### Example output https://s3.varg.ai/how-to/previews/219258f3ad0d/iart-kinetic-typography.mp4 — README showcase GIF: 'Words that move', 'Fast', 'Make it feel alive', 'Rise / Fall / Repeat' set as kinetic type (splits, mask and blur reveals, weight changes). First 6.6 s, before the iart.ai end card. (credit: iart.ai (kinetic-typography-skills), https://github.com/iart-ai/kinetic-typography-skills#readme) ##### How-to tutorials that use it - [How to make a SaaS launch video with AI](https://varg.ai/how-to/make-a-saas-launch-video.md): A polished launch film for an app or website: UI cards, device mockups, kinetic type, music. - [How to turn release notes into a changelog video with AI](https://varg.ai/how-to/turn-release-notes-into-a-changelog-video.md): Release notes, a PR or a GitHub repo turned into a short video for X or LinkedIn. - [How to make a lyric video with AI](https://varg.ai/how-to/make-a-lyric-video.md): Lyrics synced to the song: the sung line big and highlighted, karaoke-style, over art or a shader. ##### More text & titles skills - [Canopy Part Title](https://varg.ai/how-to/skills/hyperframes-canopy-part-title.md): Leaves sweep across the frame and part to reveal your headline: a 12-second 1080p title block. - [Code Slice Hero](https://varg.ai/how-to/skills/hyperframes-code-slice-hero.md): A tiled headline flips cell by cell under a sweeping depth field to reveal a second line: an 8-second hero. - [Glass Shard Title](https://varg.ai/how-to/skills/hyperframes-glass-shard-title.md): Glass shards fly in through fog and tile themselves into your headline, then fly out past the camera. - [Wireframe Portal Title](https://varg.ai/how-to/skills/hyperframes-wireframe-portal-title.md): A wireframe portal bursts open, the title comes through, then its letters swap into a second phrase. - [PixiJS Text](https://varg.ai/how-to/skills/pixijs-scene-text.md): Render and animate text in PixiJS v8: canvas Text, BitmapText, HTMLText and per-character SplitText. - [Title & Kinetic Typography Direction](https://varg.ai/how-to/skills/calesthio-title-kinetic-typography.md): Typography direction for video: text jobs, hierarchy, reveal-hold-exit timing, safe zones, contrast, font licensing. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Persian Motion Director](https://varg.ai/how-to/skills/persian-motion-director.md): Persian/RTL kinetic type done right: HarfBuzz-shaped letters, kashida, unbroken joins, font loaded before render. Persian/RTL kinetic type done right: HarfBuzz-shaped letters, kashida, unbroken joins, font loaded before render. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Text & titles](https://varg.ai/how-to/browse/skills/kinetic-type.md) - **Slug:** `persian-motion-director` - **Status:** beta - **Publisher:** atmirrr (community) - **License:** MIT - **Source:** https://github.com/atmirrr/persian-motion-director - **Works in:** Claude Code - **Requires:** Node.js with Remotion 4.0.530 (pinned in assets/remotion/package.json), Python 3 with uharfbuzz and fonttools for shape_persian.py, A native Persian reader to sign off on copy - **Cost:** Free — renders locally (Remotion needs a company license for for-profit teams above 3 employees) - **Adoption:** 5 installs, 147 GitHub stars (skills.sh, 2026-10-26) ##### What it does Directs motion design with correct Persian/Farsi and right-to-left typography in Remotion or plain web code. Its rules ban per-letter boxes and per-letter fades for joined script, pad masks for Persian ascenders and dots, set `dir`, `lang` and `` correctly, replace letter-spacing with kashida, and load the font with `FontFace` + `delayRender` so Chrome never falls back silently. Tooling covers HarfBuzz shaping to SVG outlines split into letter bodies, dots and connected groups, a zero-width-joiner splitter for per-letter animation, and a linter for Persian copy and layout code; eight technique sketches (one camera, beat cuts, line boil, dots-last, kashida) run in a browser or through Remotion. It ships Vazirmatn (OFL) with its license text. The repo is three weeks old with a single-day commit history, so it is listed as beta. ##### Use when - Motion graphics or kinetic type that contains Persian, Farsi, Arabic-script or RTL text - Persian looks broken, disconnected or clipped in a render - Per-letter or per-dot animation of Persian words - A motion piece looks like a slideshow or 'AI-made' ##### Not for - Latin-only kinetic type (use kinetic-typography) - Per-letter Nastaliq animation: the joiner split loses its cascade, animate by word or line ##### Install - Claude Code: `git clone https://github.com/atmirrr/persian-motion-director ~/.claude/skills/persian-motion-director` ##### Example output https://s3.varg.ai/how-to/previews/96a5f001fd8d/persian-motion-director.mp4 — README technique 01 'One camera, no cuts': a Persian (RTL) product promo with correctly shaped Farsi type, scenes on one strip moving right to left with whip pans and parallax. Full 7 s loop, converted from GIF. (credit: atmirrr (persian-motion-director), https://github.com/atmirrr/persian-motion-director#readme) ##### More text & titles skills - [Canopy Part Title](https://varg.ai/how-to/skills/hyperframes-canopy-part-title.md): Leaves sweep across the frame and part to reveal your headline: a 12-second 1080p title block. - [Code Slice Hero](https://varg.ai/how-to/skills/hyperframes-code-slice-hero.md): A tiled headline flips cell by cell under a sweeping depth field to reveal a second line: an 8-second hero. - [Glass Shard Title](https://varg.ai/how-to/skills/hyperframes-glass-shard-title.md): Glass shards fly in through fog and tile themselves into your headline, then fly out past the camera. - [Wireframe Portal Title](https://varg.ai/how-to/skills/hyperframes-wireframe-portal-title.md): A wireframe portal bursts open, the title comes through, then its letters swap into a second phrase. - [PixiJS Text](https://varg.ai/how-to/skills/pixijs-scene-text.md): Render and animate text in PixiJS v8: canvas Text, BitmapText, HTMLText and per-character SplitText. - [Kinetic Typography](https://varg.ai/how-to/skills/iart-kinetic-typography.md): Kinetic type for web and video: line, word or char splits, mask and blur reveals, variable-font motion, frozen-frame QA. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt ### Logo animations & intros - [Cuboid Carousel](https://varg.ai/how-to/skills/hyperframes-cuboid-carousel.md): A chain of bevelled 3D cuboids rides a travelling wave as a content carousel for features or products. A chain of bevelled 3D cuboids rides a travelling wave as a content carousel for features or products. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Logo animations & intros](https://varg.ai/how-to/browse/skills/brand-motion.md) - **Slug:** `hyperframes-cuboid-carousel` - **Status:** live - **Publisher:** HeyGen (vendor, verified) - **License:** Apache-2.0 - **Source:** https://github.com/heygen-com/hyperframes/tree/main/registry/blocks/cuboid-carousel - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** HyperFrames project (Node.js 22+, FFmpeg) - **Cost:** Free — open-source, renders locally - **Adoption:** 10,100 installs, 73,064 GitHub stars (skills.sh, 2026-10-26) ##### What it does A HyperFrames registry block: a chain of bevelled cuboids rides a travelling wave as a content carousel (1920×1080, about 6.7 s, 40 variables). Install it into a project with `npx hyperframes add cuboid-carousel`, mount it from `index.html` and set its variables; it renders locally like any other composition. ##### Use when - A ready-made title or motion moment instead of hand-built animation - Openers, chapter cards and launch-film beats in a HyperFrames project ##### Install - Claude Code: `npx skills add heygen-com/hyperframes --skill cuboid-carousel -a claude-code` - Codex: `npx skills add heygen-com/hyperframes --skill cuboid-carousel -a codex` - Cursor: `npx skills add heygen-com/hyperframes --skill cuboid-carousel -a cursor` - Gemini CLI: `npx skills add heygen-com/hyperframes --skill cuboid-carousel -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/68a80d7dd39c/hyperframes-cuboid-carousel.mp4 — Official catalog render of the block: a chain of bevelled card cuboids rides a travelling wave, then one card settles and swings to camera (credit: HeyGen (HyperFrames), https://hyperframes.heygen.com/catalog/blocks/cuboid-carousel) ##### More logo animations & intros skills - [HyperFrames Figma](https://varg.ai/how-to/skills/hyperframes-figma.md): Turn a Figma link into video: rendered assets, brand tokens, components and storyboard frames rebuilt as motion. - [Frost Sequence Camera Orbit](https://varg.ai/how-to/skills/hyperframes-frost-sequence-camera-orbit.md): An orbiting camera follows an ice logo as it shatters, reforms into two text moments and fades out. - [Krea Motion](https://varg.ai/how-to/skills/krea-motion.md): Animate products, renders and logos into reveal films and web loops with Seedance 2.5 on Krea. - [Pixel2Motion Logo Animation](https://varg.ai/how-to/skills/pixel2motion.md): Raster logo to clean SVG to a choreographed logo animation in HTML, with geometry and motion QA evidence. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Frost Sequence Camera Orbit](https://varg.ai/how-to/skills/hyperframes-frost-sequence-camera-orbit.md): An orbiting camera follows an ice logo as it shatters, reforms into two text moments and fades out. An orbiting camera follows an ice logo as it shatters, reforms into two text moments and fades out. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Logo animations & intros](https://varg.ai/how-to/browse/skills/brand-motion.md) - **Slug:** `hyperframes-frost-sequence-camera-orbit` - **Status:** live - **Publisher:** HeyGen (vendor, verified) - **License:** Apache-2.0 - **Source:** https://github.com/heygen-com/hyperframes/tree/main/registry/blocks/frost-sequence-camera-orbit - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** HyperFrames project (Node.js 22+, FFmpeg) - **Cost:** Free — open-source, renders locally - **Adoption:** 10,200 installs, 73,064 GitHub stars (skills.sh, 2026-10-26) ##### What it does A HyperFrames registry block: an orbiting camera follows an ice logo as it breaks apart, reforms into two text moments and fades (1920×1080, 22.5 s, 216 variables). Install it into a project with `npx hyperframes add frost-sequence-camera-orbit`, mount it from `index.html` and set its variables; it renders locally like any other composition. ##### Use when - A ready-made title or motion moment instead of hand-built animation - Openers, chapter cards and launch-film beats in a HyperFrames project ##### Install - Claude Code: `npx skills add heygen-com/hyperframes --skill frost-sequence-camera-orbit -a claude-code` - Codex: `npx skills add heygen-com/hyperframes --skill frost-sequence-camera-orbit -a codex` - Cursor: `npx skills add heygen-com/hyperframes --skill frost-sequence-camera-orbit -a cursor` - Gemini CLI: `npx skills add heygen-com/hyperframes --skill frost-sequence-camera-orbit -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/2570c1aa4cfd/hyperframes-frost-sequence-camera-orbit.mp4 — Official catalog render of the block, first 10 s: the ice logo shatters under an orbiting camera and reforms into 'Hard to break.' (credit: HeyGen (HyperFrames), https://hyperframes.heygen.com/catalog/blocks/frost-sequence-camera-orbit) ##### More logo animations & intros skills - [HyperFrames Figma](https://varg.ai/how-to/skills/hyperframes-figma.md): Turn a Figma link into video: rendered assets, brand tokens, components and storyboard frames rebuilt as motion. - [Cuboid Carousel](https://varg.ai/how-to/skills/hyperframes-cuboid-carousel.md): A chain of bevelled 3D cuboids rides a travelling wave as a content carousel for features or products. - [Krea Motion](https://varg.ai/how-to/skills/krea-motion.md): Animate products, renders and logos into reveal films and web loops with Seedance 2.5 on Krea. - [Pixel2Motion Logo Animation](https://varg.ai/how-to/skills/pixel2motion.md): Raster logo to clean SVG to a choreographed logo animation in HTML, with geometry and motion QA evidence. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Krea Motion](https://varg.ai/how-to/skills/krea-motion.md): Animate products, renders and logos into reveal films and web loops with Seedance 2.5 on Krea. Animate products, renders and logos into reveal films and web loops with Seedance 2.5 on Krea. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Logo animations & intros](https://varg.ai/how-to/browse/skills/brand-motion.md) - **Slug:** `krea-motion` - **Status:** live - **Publisher:** Krea (vendor, verified) - **License:** MIT - **Source:** https://github.com/krea-ai/skills/tree/main/krea-motion - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** Krea account, Krea MCP server connected - **Cost:** Billed in Krea compute units (CU) - **Adoption:** 132 installs, 29 GitHub stars (skills.sh, 2026-10-26) ##### What it does Product, brand and marketing motion on Krea with Seedance: animating a product still, render, logo or mark, reveal sequences, luxury showcases, beauty macro shots and web or product-page loops. Craft references cover cut architecture, dimensional motion and retakes. ##### Use when - Animating a product shot or logo - A reveal film or a looping web hero - Luxury or beauty macro motion ##### Not for - Anime, character or narrative animation (use Krea Generate) - Ad creative (use Krea Marketing) ##### Install - Claude Code: `/plugin marketplace add krea-ai/skills && /plugin install krea@krea` - Codex: `npx skills add krea-ai/skills --skill krea-motion -a codex` - Cursor: `npx skills add krea-ai/skills --skill krea-motion -a cursor` - Gemini CLI: `npx skills add krea-ai/skills --skill krea-motion -a gemini-cli` **Models it drives:** seedance_2_5 ##### Example output https://s3.varg.ai/how-to/previews/6f59f44ce20a/krea-motion.mp4 — 'Macro coffee pour' from Krea's Seedance 2.5 prompt gallery: a slow-motion product-beauty macro of espresso and crema, the kind of shot krea-motion builds (credit: Krea, https://www.krea.ai/models/seedance-2-5/prompt-gallery) ##### How-to tutorials that use it - [How to turn a product photo into a video with AI](https://varg.ai/how-to/turn-a-product-photo-into-a-video.md): A product shot turned into a short promo: orbit, push-in, exploded view or bullet time, with music. - [How to make a logo animation with AI](https://varg.ai/how-to/make-a-logo-animation.md): Your logo animated into a short intro or outro: 3D chrome, liquid, shader, glitch or clean motion. ##### More logo animations & intros skills - [HyperFrames Figma](https://varg.ai/how-to/skills/hyperframes-figma.md): Turn a Figma link into video: rendered assets, brand tokens, components and storyboard frames rebuilt as motion. - [Cuboid Carousel](https://varg.ai/how-to/skills/hyperframes-cuboid-carousel.md): A chain of bevelled 3D cuboids rides a travelling wave as a content carousel for features or products. - [Frost Sequence Camera Orbit](https://varg.ai/how-to/skills/hyperframes-frost-sequence-camera-orbit.md): An orbiting camera follows an ice logo as it shatters, reforms into two text moments and fades out. - [Pixel2Motion Logo Animation](https://varg.ai/how-to/skills/pixel2motion.md): Raster logo to clean SVG to a choreographed logo animation in HTML, with geometry and motion QA evidence. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Pixel2Motion Logo Animation](https://varg.ai/how-to/skills/pixel2motion.md): Raster logo to clean SVG to a choreographed logo animation in HTML, with geometry and motion QA evidence. Raster logo to clean SVG to a choreographed logo animation in HTML, with geometry and motion QA evidence. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Logo animations & intros](https://varg.ai/how-to/browse/skills/brand-motion.md) - **Slug:** `pixel2motion` - **Status:** live - **Publisher:** NolanLai (community) - **License:** MIT - **Source:** https://github.com/nolangz/pixel2motion - **Works in:** Codex, Claude Code - **Requires:** Python 3 with Pillow, NumPy and Playwright, Chromium (playwright install chromium) - **Cost:** Free — local - **Adoption:** 309 installs, 2,515 GitHub stars (GitHub + skills.sh, 2026-10-26) ##### What it does Turns a PNG, JPG, WebP or screenshot logo into a minimal SVG (edge smoothness is the hard gate, with IoU against the source pushed as high as is reasonable), then choreographs a logo reveal or idle, hover or loading motion using Disney's twelve animation principles, delivered as a standalone JS-rendered HTML showcase. Python scripts trace, overlay and audit the vector fit; Playwright scripts capture deterministic frames, probe easing curves and sweep ink continuity across handoffs. Version 2 adds self-crossing draw-on choreography and variable-width ribbon fitting. ##### Use when - Animate a logo or build a logo reveal, splash screen or brand intro - Vectorize a raster mark into motion-ready SVG ##### Not for - 3D logo stings - Character animation ##### Install - Codex: `npx skills add nolangz/pixel2motion` - Claude Code: `npx skills add nolangz/pixel2motion` ##### Example output https://s3.varg.ai/how-to/previews/5492fe729b30/pixel2motion.mp4 — README showcase 'Horizon': the motion output the skill built from a raster logo (sun, horizon lines and wordmark drawn in). Full 2 s loop, converted from GIF. (credit: NolanLai (pixel2motion), https://github.com/nolangz/pixel2motion#readme) ##### How-to tutorials that use it - [How to make a logo animation with AI](https://varg.ai/how-to/make-a-logo-animation.md): Your logo animated into a short intro or outro: 3D chrome, liquid, shader, glitch or clean motion. ##### More logo animations & intros skills - [HyperFrames Figma](https://varg.ai/how-to/skills/hyperframes-figma.md): Turn a Figma link into video: rendered assets, brand tokens, components and storyboard frames rebuilt as motion. - [Cuboid Carousel](https://varg.ai/how-to/skills/hyperframes-cuboid-carousel.md): A chain of bevelled 3D cuboids rides a travelling wave as a content carousel for features or products. - [Frost Sequence Camera Orbit](https://varg.ai/how-to/skills/hyperframes-frost-sequence-camera-orbit.md): An orbiting camera follows an ice logo as it shatters, reforms into two text moments and fades out. - [Krea Motion](https://varg.ai/how-to/skills/krea-motion.md): Animate products, renders and logos into reveal films and web loops with Seedance 2.5 on Krea. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt ### 3D scenes - [Three.js Architecture Effects](https://varg.ai/how-to/skills/threejs-architecture-effects.md): Self-assembling 3D buildings in three.js: bricks, timber, eaves and PBR, driven by one seekable progress value. Self-assembling 3D buildings in three.js: bricks, timber, eaves and PBR, driven by one seekable progress value. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [3D scenes](https://varg.ai/how-to/browse/skills/3d-scenes.md) - **Slug:** `threejs-architecture-effects` - **Status:** beta - **Publisher:** Hailey (community) - **License:** MIT - **Source:** https://github.com/lhlGitHub/threejs-architecture-effects - **Works in:** Claude Code, Codex, Cursor - **Requires:** Node.js 22.13+ and npm, A WebGL2 browser - **Cost:** Free — open-source starter, runs locally - **Adoption:** 6 installs, 631 GitHub stars (GitHub + skills.sh, 2026-10-26) ##### What it does Builds orbitable three.js construction animations, mainly Chinese pavilions, pagodas and clock towers that assemble brick by brick. The order runs from foundations and timber frame to layered eaves, tiles and ornament, with procedural PBR materials and close-up detail cameras. One normalized progress value, shared by the color and depth shaders, drives the whole build, so seeking backward, pausing and replaying stay deterministic. That makes it easy to drive from a video timeline. `scripts/scaffold.mjs` copies a Vite + React + three r185 starter (a clock-tower reference) into a new folder without installing anything. It is beta because the repo was published in one day (2026-09-22, three commits). The starter is interactive (orbit, scrub bar), throttles shadow updates on wall-clock time and skips frames while the page is hidden, so a capture loop should set `progress` and force a render on every frame. ##### Use when - Brick-by-brick or storey-by-storey building growth animation - Chinese pavilions, pagodas or clock towers assembled in 3D - Architectural miniatures with orbit and detail cameras ##### Not for - Flat image reveals or generic website styling ##### Install - Claude Code: `npx skills add https://github.com/lhlgithub/threejs-architecture-effects --skill threejs-architecture-effects` - Cursor: `git clone https://github.com/lhlGitHub/threejs-architecture-effects.git && cp -R threejs-architecture-effects ~/.cursor/skills/` ##### Example output https://s3.varg.ai/how-to/previews/842cc4af5987/threejs-architecture-effects.mp4 — The README preview video: a Chinese drum tower assembles itself in three.js, from empty lot to timber frame, red walls and tiered green roofs, driven by one progress value. Cut 1.5-11.5 s of 19.6 s. (credit: Hailey (lhlGitHub, threejs-architecture-effects), https://github.com/lhlGitHub/threejs-architecture-effects#readme) ##### More 3D scenes skills - [Three.js Skills](https://varg.ai/how-to/skills/threejs-skills.md): 13 Three.js skills: scenes, geometry, materials, lighting, GLSL and TSL shaders, animation, post-processing, R3F. - [Three.js Scene Composition](https://varg.ai/how-to/skills/calesthio-threejs-scene-composition.md): Plan, build and capture Three.js scenes for video: absolute-time frames, seeded motion, color-safe glTF, render QA. - [R3F Skills](https://varg.ai/how-to/skills/enzed-r3f-skills.md): 11 React Three Fiber skills checked against Fiber 9.7 and three r185: Canvas, demand loops, loaders, shaders, post. - [Procedural 3D Story](https://varg.ai/how-to/skills/karpathy-3d-story.md): Turn a myth or book into a seeded, multi-chapter procedural three.js film in one scene.html, audited frame by frame. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt ### Phone & app screens - [iPhone Duo Fold](https://varg.ai/how-to/skills/iphone-duo-fold.md): Your screens on Apple's foldable iPhone Duo, unfolding in 3D with its progressive blur. Local, free. Your screens on Apple's foldable iPhone Duo, unfolding in 3D with its progressive blur. Local, free. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Phone & app screens](https://varg.ai/how-to/browse/skills/native-ui.md) - **Slug:** `iphone-duo-fold` - **Status:** live - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/code2video-templates/tree/master/iphone-duo-fold - **Works in:** Claude Code, Codex, Cursor - **Requires:** bun, ffmpeg, Google Chrome or Chromium - **Cost:** Free — renders locally, no varg API key - **Runtime:** ~20 s per 6 s clip at 1080p ##### What it does Apple's foldable iPhone in 3D, folding open or shut with your screens: the cover shows while it's folded, the inner screen when it opens, and the moving half blurs and darkens toward its edge exactly like the launch film. A screen can be an image, a screen recording, an HTML page the agent writes for you, or a URL, and one picture can span both screens so the folded view continues into the open one. Four presets (`unfold`, `fold`, `loop`, `reveal`) or your own keyframes for the hinge and the camera, Star White or Night Sky, a caption that sits with the phone at any aspect, MP4 at any size or transparent ProRes 4444 / WebM for an edit. Frames are rendered one by one in headless Chrome and piped into ffmpeg, so the same flags always give the same video; a 6 s 1080p clip takes about 20 s. The fold deformation and blur shaders are ported from [jadon7/iphone-duo](https://github.com/jadon7/iphone-duo) (MIT); Apple's AR model is fetched from apple.com on the first run and never bundled. ##### Use when - An iPhone Duo or foldable iPhone mockup video, an app promo on a folding phone - An unfold reveal, a fold animation or a fold-unfold loop with your own screens - A transparent phone clip to drop into an edit - Requests like "айфон дуо", "складной айфон", "анимация складывания", "покажи приложение на iPhone Duo" ##### Not for - Flat phone mockups or screen swipes (use stamp-carousel or apple-notes-video) - Generated live-action footage of a phone in someone's hands (use varg-ai) ##### Install - Claude Code: `git clone --depth 1 https://github.com/vargHQ/code2video-templates ~/code2video-templates && ln -s ~/code2video-templates/iphone-duo-fold ~/.claude/skills/iphone-duo-fold` ##### Example output https://s3.varg.ai/how-to/previews/b1adcaedb8e5/iphone-duo-fold.mp4 — Rendered with the template itself: bun render.ts --config examples/aurora.json (a generated panorama across both screens; folded, the cover shows the cabin and the clock, unfolding reveals the lake). 1920x1080, 7 s. (credit: varg, https://github.com/vargHQ/code2video-templates/tree/master/iphone-duo-fold) ##### More phone & app screens skills - [iMessage Video](https://varg.ai/how-to/skills/varg-imessage-video.md): Pixel-perfect iMessage conversation videos from a JSON script: iOS 26 chrome, typing, tapbacks, group chats. - [Duo Foldable Phone](https://varg.ai/how-to/skills/hyperframes-duo.md): Two HTML screens composited into a photo of hands holding an open foldable phone, for side-by-side meme videos. - [X Posting License](https://varg.ai/how-to/skills/hyperframes-x-posting-license.md): A 10-second animated 'posting license' card for an X profile, filled with real profile data into a locked template. - [App Store Screenshots](https://varg.ai/how-to/skills/inference-sh-app-store-screenshots.md): Make App Store and Google Play screenshots to exact specs: sizes, gallery order, mockups and previews. - [Gooseworks Video (Apple Notes Ads)](https://varg.ai/how-to/skills/gooseworks-apple-notes-ads.md): Video ads in chat from a reference ad or a ready style, including the viral HTML-only Apple Notes typing ad. - [Duo Foldable Phone](https://varg.ai/how-to/skills/hyperframes-duo.md): Two HTML screens composited into a photo of hands holding an open foldable phone, for side-by-side meme videos. Two HTML screens composited into a photo of hands holding an open foldable phone, for side-by-side meme videos. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Phone & app screens](https://varg.ai/how-to/browse/skills/native-ui.md) - **Slug:** `hyperframes-duo` - **Status:** beta - **Publisher:** HeyGen (vendor, verified) - **License:** Apache-2.0 - **Source:** https://github.com/heygen-com/hyperframes-community-skills/tree/master/skills/duo - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** Node.js 22+ and FFmpeg for the HyperFrames CLI (npx) - **Cost:** Free — open-source, renders locally - **Adoption:** 196 installs, 231 GitHub stars (skills.sh, 2026-10-26) ##### What it does Composite any two HTML screens into a locked photo plate of hands holding an open foldable phone — a 1448×1086 still with two blank 495×849 / 498×849 screen slots (left and right) — to make side-by-side meme and comparison videos: TikTok vs Reels, ChatGPT vs Claude, "me vs my friend", before/after, two apps racing. Trigger on: "foldable phone meme", "put X on the left screen and Y on the right", "two-screen comparison video", "the dual-screen phone meme", or any request that names two apps/feeds/chats to show at once. The plate, geometry, and camera are finished — the agent only authors what plays INSIDE the two screens and renders. ##### Use when - Two apps, feeds or chats side by side on one phone - "X vs Y" comparison memes - Before/after or two versions racing ##### Not for - A single phone, a laptop or a device photo you supply - Other aspect ratios (crop the 4:3 render as a follow-up) ##### Install - Claude Code: `npx skills add heygen-com/hyperframes-community-skills --skill duo -a claude-code` - Codex: `npx skills add heygen-com/hyperframes-community-skills --skill duo -a codex` - Cursor: `npx skills add heygen-com/hyperframes-community-skills --skill duo -a cursor` - Gemini CLI: `npx skills add heygen-com/hyperframes-community-skills --skill duo -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/606355b9f5cc/hyperframes-duo.mp4 — The skill's 1448x1086 foldable-phone plate with two live screens side by side (a TikTok feed vs a Reels feed); linked as the duo skill demo (credit: Jake Moran (@jake11moran), HeyGen, https://x.com/jake11moran/status/2098118454343798862) ##### How-to tutorials that use it - [How to make an app demo video with AI](https://varg.ai/how-to/make-an-app-demo-video.md): A polished demo of your app: device mockups, real screens, captions, music — for the App Store, ads or a landing page. ##### More phone & app screens skills - [iPhone Duo Fold](https://varg.ai/how-to/skills/iphone-duo-fold.md): Your screens on Apple's foldable iPhone Duo, unfolding in 3D with its progressive blur. Local, free. - [iMessage Video](https://varg.ai/how-to/skills/varg-imessage-video.md): Pixel-perfect iMessage conversation videos from a JSON script: iOS 26 chrome, typing, tapbacks, group chats. - [X Posting License](https://varg.ai/how-to/skills/hyperframes-x-posting-license.md): A 10-second animated 'posting license' card for an X profile, filled with real profile data into a locked template. - [App Store Screenshots](https://varg.ai/how-to/skills/inference-sh-app-store-screenshots.md): Make App Store and Google Play screenshots to exact specs: sizes, gallery order, mockups and previews. - [Gooseworks Video (Apple Notes Ads)](https://varg.ai/how-to/skills/gooseworks-apple-notes-ads.md): Video ads in chat from a reference ad or a ready style, including the viral HTML-only Apple Notes typing ad. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [X Posting License](https://varg.ai/how-to/skills/hyperframes-x-posting-license.md): A 10-second animated 'posting license' card for an X profile, filled with real profile data into a locked template. A 10-second animated 'posting license' card for an X profile, filled with real profile data into a locked template. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Phone & app screens](https://varg.ai/how-to/browse/skills/native-ui.md) - **Slug:** `hyperframes-x-posting-license` - **Status:** beta - **Publisher:** HeyGen (vendor, verified) - **License:** Apache-2.0 - **Source:** https://github.com/heygen-com/hyperframes-community-skills/tree/master/skills/x-posting-license - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** Node.js 18+ and the HyperFrames CLI via npx, FFmpeg, A browser tool to read the X profile, macOS GPU rendering for the WebGL dot field - **Cost:** Free — open-source, renders locally - **Adoption:** 214 installs, 231 GitHub stars (skills.sh, 2026-10-26) ##### What it does Turn any X (Twitter) profile into a 10.3s animated "POSTING LICENSE" card video — a fake driver's license in X's branding: the card slides in over an animated dot-matrix ground, license text types on letter-by-letter, follower / following / post counters tick up, a diagonal glimmer sweeps as the card tilts in 3D, it flips to a fine-print back with a signature write-on, then crouches and leaps out of frame. Trigger on: "X driver's license", "posting license", "make an X license for @handle", "license card video". The whole composition is a finished template — the agent only supplies profile data and renders. ##### Use when - An X 'posting license' or driver's license card for a handle - A short profile card video from real X data ##### Not for - Custom card designs or edited motion (the template is locked; use HyperFrames) - Invented counts or bios ##### Install - Claude Code: `npx skills add heygen-com/hyperframes-community-skills --skill x-posting-license -a claude-code` - Codex: `npx skills add heygen-com/hyperframes-community-skills --skill x-posting-license -a codex` - Cursor: `npx skills add heygen-com/hyperframes-community-skills --skill x-posting-license -a cursor` - Gemini CLI: `npx skills add heygen-com/hyperframes-community-skills --skill x-posting-license -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/b46762a3033d/hyperframes-x-posting-license.mp4 — Skill launch example: the animated X 'posting license' ID card filled with the author's own profile data, flipping to its terms on the back (credit: Jake Moran (@jake11moran), skill author at HeyGen, https://x.com/jake11moran/status/2093056828410397064) ##### More phone & app screens skills - [iPhone Duo Fold](https://varg.ai/how-to/skills/iphone-duo-fold.md): Your screens on Apple's foldable iPhone Duo, unfolding in 3D with its progressive blur. Local, free. - [iMessage Video](https://varg.ai/how-to/skills/varg-imessage-video.md): Pixel-perfect iMessage conversation videos from a JSON script: iOS 26 chrome, typing, tapbacks, group chats. - [Duo Foldable Phone](https://varg.ai/how-to/skills/hyperframes-duo.md): Two HTML screens composited into a photo of hands holding an open foldable phone, for side-by-side meme videos. - [App Store Screenshots](https://varg.ai/how-to/skills/inference-sh-app-store-screenshots.md): Make App Store and Google Play screenshots to exact specs: sizes, gallery order, mockups and previews. - [Gooseworks Video (Apple Notes Ads)](https://varg.ai/how-to/skills/gooseworks-apple-notes-ads.md): Video ads in chat from a reference ad or a ready style, including the viral HTML-only Apple Notes typing ad. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt ### Editing & clips - [Clipify](https://varg.ai/how-to/skills/clipify.md): Long videos into social clips: find funny lines or impact beats, cut, reframe to 9:16, burn word-by-word captions. Long videos into social clips: find funny lines or impact beats, cut, reframe to 9:16, burn word-by-word captions. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Editing & clips](https://varg.ai/how-to/browse/skills/editing.md) - **Slug:** `clipify` - **Status:** live - **Publisher:** Louise de Sadeleer (community) - **License:** MIT - **Source:** https://github.com/louisedesadeleer/clipify - **Works in:** Claude Code - **Requires:** ffmpeg with libx264, OpenAI Whisper (openai-whisper), Python 3 with NumPy - **Cost:** Free — local - **Adoption:** 176 installs, 611 GitHub stars (GitHub + skills.sh, 2026-10-26) ##### What it does Finds compelling moments in a video, either funny dialogue (from a Whisper transcript) or repeated impact actions such as axe chops, hits, throws or drumbeats (from audio transient detection), checks that each candidate is in frame, and cuts them as standalone clips. It can reformat 16:9 to 9:16 with a pan between faces or a split screen, time-warp pans for tighter reveals, and burn opus-style word-by-word captions from a fresh transcript of each trimmed clip. Everything runs locally with ffmpeg, Whisper and Python. ##### Use when - Make Shorts or Reels from a long video - Cut an action montage on every hit or chop - Reframe 16:9 footage to vertical with captions ##### Not for - Generating new footage - Multi-camera edits ##### Install - Claude Code: `git clone https://github.com/louisedesadeleer/clipify.git ~/.claude/skills/clipify` **Models it drives:** whisper ##### Example output https://s3.varg.ai/how-to/previews/76ea06e94e22/clipify.mp4 — From the creator's demo post: a 9:16 clip that /clipify cut from her long video (funny moment, face-tracked reframe, burned word-by-word captions) plays beside her webcam. Cut 112-120 s of 148 s. (credit: Louise de Sadeleer (clipify), https://x.com/LouiseDSadeleer/status/2051652715617378655) ##### How-to tutorials that use it - [How to cut a podcast into short video clips with AI](https://varg.ai/how-to/cut-a-podcast-into-short-video-clips.md): Find the best moments in a long video, cut them, reframe to 9:16 and burn captions. - [How to edit your own video footage with AI](https://varg.ai/how-to/edit-your-own-video-footage.md): Cut silences and flubs, reframe 16:9 to 9:16, add captions and b-roll to footage you shot. ##### More editing & clips skills - [Short-form FFmpeg cookbook: copyable commands for Reels](https://varg.ai/how-to/skills/short-form-ffmpeg-cookbook.md): Footage into a 9:16 MP4 with FFmpeg: crop to vertical, burn captions, trim and join, music under the voice. - [Splice AI clips into a video with FFmpeg](https://varg.ai/how-to/skills/varg-ffmpeg-mixing.md): Splice generated clips into an original video with ffmpeg without A/V drift: absolute trims, one audio track. - [Chengfeng Videocut](https://varg.ai/how-to/skills/chengfeng-videocut.md): Chinese talking-head editing: word-level transcript, five-pass flub sweep, human review in a Studio, subtitles, export. - [Claude Code Video Toolkit](https://varg.ai/how-to/skills/claude-code-video-toolkit.md): Open-source video workspace for Claude Code: script, voice, music, visuals and Remotion render on your own GPU account. - [Herk Edit Kit](https://varg.ai/how-to/skills/herk-edit-kit.md): Nate Herk's kit to edit talking-head videos, reels and Shorts: silence and flub cuts, storytelling, motion graphics. - [VectCut](https://varg.ai/how-to/skills/vectcut.md): Drives the VectCut editing API: talking-head cuts, remixes, e-commerce ads, subtitles, effects, cloud render ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt ### Captions & subtitles - [3D Camera Captions](https://varg.ai/how-to/skills/hyperframes-camera-3d-captions.md): Captions that live in 3D around a talking head: the camera flies through the words, heroes pass behind the speaker. Captions that live in 3D around a talking head: the camera flies through the words, heroes pass behind the speaker. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Captions & subtitles](https://varg.ai/how-to/browse/skills/captions.md) - **Slug:** `hyperframes-camera-3d-captions` - **Status:** live - **Publisher:** HeyGen (vendor, verified) - **License:** Apache-2.0 - **Source:** https://github.com/heygen-com/hyperframes-community-skills/tree/master/skills/camera-3d-captions - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** Node.js 22+, FFmpeg, Python 3 with fonttools, brotli, numpy and pillow, A locked-off talking-head take and word timings from your own transcriber, p5 Paint Animation skill (optional, for hand-drawn words) - **Cost:** Free — open-source, renders locally; the person matte runs on-device - **Adoption:** 264 installs, 231 GitHub stars (skills.sh, 2026-10-26) ##### What it does Rebuilds the After Effects 3D-camera caption look as a deterministic per-frame painter for HyperFrames: caption groups sit at different depths so camera moves pull them apart, hero words hide behind the speaker through an alpha matte, and a ring of words turns in front. Prep scripts cut the take, build a reflect-padded plate and a local person matte with `hyperframes remove-background`, and measure font metrics; in-browser checks catch words leaving the frame or touching. It needs a locked-off talking-head shot and word timings from a transcriber you already run. A community-tier skill written by HeyGen's Jake Moran. ##### Use when - 3D text or 3D camera captions on a talking-head video - Words placed around or behind the speaker - A camera move that flies through the captions - Depth captions on an avatar video ##### Not for - Plain subtitles (use HyperFrames) - Moving or handheld shots, extreme close-ups, or subjects a matting model can't separate ##### Install - Claude Code: `npx skills add heygen-com/hyperframes-community-skills --skill camera-3d-captions -a claude-code` - Codex: `npx skills add heygen-com/hyperframes-community-skills --skill camera-3d-captions -a codex` - Cursor: `npx skills add heygen-com/hyperframes-community-skills --skill camera-3d-captions -a cursor` - Gemini CLI: `npx skills add heygen-com/hyperframes-community-skills --skill camera-3d-captions -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/1fc0866a8618/hyperframes-camera-3d-captions.mp4 — Skill launch demo: captions placed in 3D around a talking head; 'forget' flies through, words pass behind the speaker and curve around her (credit: Jake Moran (@jake11moran), skill author at HeyGen, https://x.com/jake11moran/status/2102544324445848031) ##### How-to tutorials that use it - [How to cut a podcast into short video clips with AI](https://varg.ai/how-to/cut-a-podcast-into-short-video-clips.md): Find the best moments in a long video, cut them, reframe to 9:16 and burn captions. - [How to add animated captions to a video with AI](https://varg.ai/how-to/add-animated-captions-to-a-video.md): TikTok-style subtitles burned into an existing video: word highlight, pop words, or designed editorial captions. ##### More captions & subtitles skills - [Remotion Captions](https://varg.ai/how-to/skills/remotion-captions.md): Transcribe speech and burn in animated captions with Remotion, or import an existing SRT file. - [ElevenLabs Speech to Text](https://varg.ai/how-to/skills/elevenlabs-speech-to-text.md): Transcribe audio and video with word timestamps, speaker labels and keyterms, ready for subtitles. - [ElevenLabs Speech to Text (inference.sh)](https://varg.ai/how-to/skills/inference-sh-elevenlabs-stt.md): Transcribe audio with ElevenLabs Scribe, speaker diarization and forced alignment, via inference.sh. - [Speech to Text](https://varg.ai/how-to/skills/inference-sh-speech-to-text.md): Transcribe audio with ElevenLabs Scribe v2 or Whisper V3 Large, with timestamps and speaker labels. - [Pretext Text Metrics](https://varg.ai/how-to/skills/pretext-yaniv-golan.md): Measure and fit text without the DOM using pretext: auto-fit sizes, know line breaks before render, canvas/WebGL text. - [Chinese Font Selector](https://varg.ai/how-to/skills/chinese-font-selector.md): Pick Chinese fonts by license tier first (OFL, vendor-free, avoid), then by scene and mood, with Latin pairings. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Remotion Captions](https://varg.ai/how-to/skills/remotion-captions.md): Transcribe speech and burn in animated captions with Remotion, or import an existing SRT file. Transcribe speech and burn in animated captions with Remotion, or import an existing SRT file. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Captions & subtitles](https://varg.ai/how-to/browse/skills/captions.md) - **Slug:** `remotion-captions` - **Status:** live - **Publisher:** Remotion (vendor, verified) - **License:** MIT - **Source:** https://github.com/remotion-dev/skills/tree/main/skills/remotion-captions - **Works in:** Claude Code, Codex, Cursor, Gemini CLI - **Requires:** Node.js, Remotion 4 - **Cost:** Free — local render (Remotion company license for larger for-profit teams) - **Adoption:** 216,400 installs, 5,750 GitHub stars (skills.sh, 2026-10-26) ##### What it does Covers the three caption paths in Remotion: generating captions by transcribing audio, displaying and animating them on screen, and importing existing `.srt` files. Each path has its own reference file, and a post-creation section explains how to adjust captions once they exist. ##### Use when - Adding subtitles or animated captions to a Remotion video - Transcribing a voiceover into captions - Importing an SRT file ##### Install - Claude Code: `npx skills add remotion-dev/skills --skill remotion-captions -a claude-code` - Codex: `npx skills add remotion-dev/skills --skill remotion-captions -a codex` - Cursor: `npx skills add remotion-dev/skills --skill remotion-captions -a cursor` - Gemini CLI: `npx skills add remotion-dev/skills --skill remotion-captions -a gemini-cli` ##### Example output https://s3.varg.ai/how-to/previews/3953d368f7d3/remotion-captions.mp4 — Official preview of Remotion's TikTok template: speech transcribed with Whisper.cpp and burned in as animated word-by-word captions (credit: Remotion, https://www.remotion.dev/templates/tiktok) ##### How-to tutorials that use it - [How to make a UGC ad video with AI](https://varg.ai/how-to/make-a-ugc-ad-video.md): A creator-style ad: a person talks to camera about your product, with b-roll, captions and an end card. - [How to cut a podcast into short video clips with AI](https://varg.ai/how-to/cut-a-podcast-into-short-video-clips.md): Find the best moments in a long video, cut them, reframe to 9:16 and burn captions. - [How to add animated captions to a video with AI](https://varg.ai/how-to/add-animated-captions-to-a-video.md): TikTok-style subtitles burned into an existing video: word highlight, pop words, or designed editorial captions. - [How to make a lyric video with AI](https://varg.ai/how-to/make-a-lyric-video.md): Lyrics synced to the song: the sung line big and highlighted, karaoke-style, over art or a shader. ##### More captions & subtitles skills - [3D Camera Captions](https://varg.ai/how-to/skills/hyperframes-camera-3d-captions.md): Captions that live in 3D around a talking head: the camera flies through the words, heroes pass behind the speaker. - [ElevenLabs Speech to Text](https://varg.ai/how-to/skills/elevenlabs-speech-to-text.md): Transcribe audio and video with word timestamps, speaker labels and keyterms, ready for subtitles. - [ElevenLabs Speech to Text (inference.sh)](https://varg.ai/how-to/skills/inference-sh-elevenlabs-stt.md): Transcribe audio with ElevenLabs Scribe, speaker diarization and forced alignment, via inference.sh. - [Speech to Text](https://varg.ai/how-to/skills/inference-sh-speech-to-text.md): Transcribe audio with ElevenLabs Scribe v2 or Whisper V3 Large, with timestamps and speaker labels. - [Pretext Text Metrics](https://varg.ai/how-to/skills/pretext-yaniv-golan.md): Measure and fit text without the DOM using pretext: auto-fit sizes, know line breaks before render, canvas/WebGL text. - [Chinese Font Selector](https://varg.ai/how-to/skills/chinese-font-selector.md): Pick Chinese fonts by license tier first (OFL, vendor-free, avoid), then by scene and mood, with Latin pairings. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Cinematic Caption](https://varg.ai/how-to/skills/cinematic-caption-audrey.md): Designed editorial captions for HyperFrames: scored hero words, subject-aware depth, glass type — not a subtitle strip. Designed editorial captions for HyperFrames: scored hero words, subject-aware depth, glass type — not a subtitle strip. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Captions & subtitles](https://varg.ai/how-to/browse/skills/captions.md) - **Slug:** `cinematic-caption-audrey` - **Status:** beta - **Publisher:** Audrey (community) - **License:** MIT - **Source:** https://github.com/audrey-560/hyperframes-cinematic-caption - **Works in:** Claude Code, Codex - **Requires:** A HyperFrames project and the HyperFrames skills (heygen-com/hyperframes), Node.js for the HyperFrames CLI, A subject matte or cutout for behind-the-speaker words (optional) - **Cost:** Free — instructions only; renders locally with HyperFrames - **Adoption:** 17 installs, 63 GitHub stars (GitHub + skills.sh, 2026-10-26) ##### What it does Turns a transcript (SRT/VTT or word-level JSON) into designed caption moments inside an existing HyperFrames composition: it scores hero words by meaning and proof value, builds ordered mixed-case stacks, anchors type around the speaker and only layers words behind a person when a real matte exists. It writes `cinematic-caption-plan.json` first, then a namespaced overlay track or caption sub-composition on one seek-safe timeline, and verifies with `hyperframes check`, cue snapshots and a chronological contact sheet before render. Instructions only (no scripts); it leans on the HyperFrames skills and CLI for transcription, linting and preview. Young repo: three commits over two days in August 2026. ##### Use when - Cinematic or editorial captions on a HyperFrames video - Make key words, numbers or locations large and animated around a speaker - Creator-reel or real-estate captions without a fixed subtitle band ##### Not for - A plain fixed subtitle strip or bottom-center karaoke pills (the skill avoids them) - Remotion projects (use remotion-dev/remotion-captions) ##### Install - Claude Code: `git clone https://github.com/audrey-560/hyperframes-cinematic-caption.git ~/.claude/skills/cinematic-caption` - Codex: `git clone https://github.com/audrey-560/hyperframes-cinematic-caption.git ~/.agents/skills/cinematic-caption` ##### How-to tutorials that use it - [How to add animated captions to a video with AI](https://varg.ai/how-to/add-animated-captions-to-a-video.md): TikTok-style subtitles burned into an existing video: word highlight, pop words, or designed editorial captions. ##### More captions & subtitles skills - [3D Camera Captions](https://varg.ai/how-to/skills/hyperframes-camera-3d-captions.md): Captions that live in 3D around a talking head: the camera flies through the words, heroes pass behind the speaker. - [Remotion Captions](https://varg.ai/how-to/skills/remotion-captions.md): Transcribe speech and burn in animated captions with Remotion, or import an existing SRT file. - [ElevenLabs Speech to Text](https://varg.ai/how-to/skills/elevenlabs-speech-to-text.md): Transcribe audio and video with word timestamps, speaker labels and keyterms, ready for subtitles. - [ElevenLabs Speech to Text (inference.sh)](https://varg.ai/how-to/skills/inference-sh-elevenlabs-stt.md): Transcribe audio with ElevenLabs Scribe, speaker diarization and forced alignment, via inference.sh. - [Speech to Text](https://varg.ai/how-to/skills/inference-sh-speech-to-text.md): Transcribe audio with ElevenLabs Scribe v2 or Whisper V3 Large, with timestamps and speaker labels. - [Pretext Text Metrics](https://varg.ai/how-to/skills/pretext-yaniv-golan.md): Measure and fit text without the DOM using pretext: auto-fit sizes, know line breaks before render, canvas/WebGL text. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Caption Animation](https://varg.ai/how-to/skills/iart-caption-animation.md): Word-timed TikTok captions in Remotion or CSS: per-word pop, active-word highlight, 9:16 safe areas, sidecar SRT/VTT. Word-timed TikTok captions in Remotion or CSS: per-word pop, active-word highlight, 9:16 safe areas, sidecar SRT/VTT. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Captions & subtitles](https://varg.ai/how-to/browse/skills/captions.md) - **Slug:** `iart-caption-animation` - **Status:** live - **Publisher:** iart.ai (community) - **License:** MIT - **Source:** https://github.com/iart-ai/tiktok-video-skills/tree/main/skills/caption-animation - **Works in:** Claude Code, Cursor, Codex - **Requires:** Node.js and Remotion (company license for for-profit teams above 3 employees), whisper.cpp via @remotion/install-whisper-cpp, or your own cloud STT key, ffmpeg/ffprobe for the verify scripts - **Cost:** Free — renders locally with Remotion; a cloud STT, if chosen, bills your own key - **Adoption:** 1,994 installs, 17 GitHub stars (GitHub + skills.sh, 2026-10-26) ##### What it does A how-to skill for word-by-word captions on vertical video. It normalizes Whisper output (`@remotion/install-whisper-cpp` with token-level timestamps, or a cloud STT) or an existing SRT into `{text, startMs, endMs}` tokens, pages them 1–4 words at a time with `createTikTokStyleCaptions()`, and animates a spring pop plus an active-word highlight in Remotion, with a plain CSS variant for web players. The reference adds an SRT parser, an SRT/VTT emitter, per-platform safe-area numbers and a stills-first verify loop, and it refuses to fake word timing by splitting lines evenly. It builds on Remotion's own caption packages, so it overlaps remotion-dev/remotion-captions; the files end with promotional links to the publisher's hosted iart.ai agent. ##### Use when - Add animated TikTok or karaoke captions - Turn an SRT or Whisper transcript into word-by-word animated text - Sync captions to a voiceover and burn them in - Ship burned-in captions plus an accessible SRT/VTT ##### Not for - Interpolating word times from sentence-level cues (it re-transcribes instead) - HyperFrames compositions (use heygen-com/hyperframes or cinematic-caption) ##### Install - Claude Code: `npx skills add https://github.com/iart-ai/tiktok-video-skills --skill caption-animation` - Codex: `npx skills add https://github.com/iart-ai/tiktok-video-skills --skill caption-animation` - Cursor: `npx skills add https://github.com/iart-ai/tiktok-video-skills --skill caption-animation` **Models it drives:** whisper ##### How-to tutorials that use it - [How to make a UGC ad video with AI](https://varg.ai/how-to/make-a-ugc-ad-video.md): A creator-style ad: a person talks to camera about your product, with b-roll, captions and an end card. - [How to cut a podcast into short video clips with AI](https://varg.ai/how-to/cut-a-podcast-into-short-video-clips.md): Find the best moments in a long video, cut them, reframe to 9:16 and burn captions. - [How to add animated captions to a video with AI](https://varg.ai/how-to/add-animated-captions-to-a-video.md): TikTok-style subtitles burned into an existing video: word highlight, pop words, or designed editorial captions. - [How to make a faceless YouTube video with AI](https://varg.ai/how-to/make-a-faceless-youtube-video.md): A narrated short or long video with no presenter: script, voiceover, visuals, captions, music. - [How to edit your own video footage with AI](https://varg.ai/how-to/edit-your-own-video-footage.md): Cut silences and flubs, reframe 16:9 to 9:16, add captions and b-roll to footage you shot. ##### More captions & subtitles skills - [3D Camera Captions](https://varg.ai/how-to/skills/hyperframes-camera-3d-captions.md): Captions that live in 3D around a talking head: the camera flies through the words, heroes pass behind the speaker. - [Remotion Captions](https://varg.ai/how-to/skills/remotion-captions.md): Transcribe speech and burn in animated captions with Remotion, or import an existing SRT file. - [ElevenLabs Speech to Text](https://varg.ai/how-to/skills/elevenlabs-speech-to-text.md): Transcribe audio and video with word timestamps, speaker labels and keyterms, ready for subtitles. - [ElevenLabs Speech to Text (inference.sh)](https://varg.ai/how-to/skills/inference-sh-elevenlabs-stt.md): Transcribe audio with ElevenLabs Scribe, speaker diarization and forced alignment, via inference.sh. - [Speech to Text](https://varg.ai/how-to/skills/inference-sh-speech-to-text.md): Transcribe audio with ElevenLabs Scribe v2 or Whisper V3 Large, with timestamps and speaker labels. - [Pretext Text Metrics](https://varg.ai/how-to/skills/pretext-yaniv-golan.md): Measure and fit text without the DOM using pretext: auto-fit sizes, know line breaks before render, canvas/WebGL text. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [Agent Caption](https://varg.ai/how-to/skills/agent-caption.md): Local captions in any language: Whisper words, forced-aligned timing, SRT/VTT, burned styles or a GSAP captions.js. Local captions in any language: Whisper words, forced-aligned timing, SRT/VTT, burned styles or a GSAP captions.js. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Captions & subtitles](https://varg.ai/how-to/browse/skills/captions.md) - **Slug:** `agent-caption` - **Status:** beta - **Publisher:** Ali Hamza Kamboh (community) - **License:** MIT - **Source:** https://github.com/ahkamboh/agent-caption - **Works in:** Claude Code, Cursor, Codex - **Requires:** Python 3.10+, ffmpeg (libass optional), About 4 GB of disk for models, Java 8+ only for --grammar - **Cost:** Free — runs locally; models download once - **Adoption:** 41 GitHub stars (GitHub, 2026-10-26) ##### What it does A Python toolkit an agent drives from SKILL.md: Whisper large-v3 (faster-whisper) writes the words, WhisperX or Meta's MMS_FA aligner times every word on the waveform, and `caption.py` burns one of 15 styles with ffmpeg (falling back to Pillow overlays when ffmpeg lacks libass) or just writes `.srt`/`.vtt`. `scripts/caption.py` emits a `captions.js` layer (word, line or karaoke presets such as hormozi, pill, neon) that mounts on a GSAP timeline in an HTML composition, and Argos Translate makes offline translated subtitle tracks. Extras include `--script` alignment of known lyrics, Demucs vocal isolation for songs, a glossary prompt and a deterministic QA report. Setup is heavy (torch, WhisperX and about 4 GB of models) and Hinglish/code-switched speech is its focus. ##### Use when - Caption a video or song in any language, fully offline - Word timing from forced alignment instead of raw ASR timestamps - Hinglish or code-switched captions - Animated word or karaoke captions for an HTML/GSAP composition - Export SRT/VTT and translated subtitle tracks ##### Not for - Hand-timed captions or lead/lag offsets (the skill forbids them) - Cloud transcription (use elevenlabs/speech-to-text or Gladia) ##### Install - Claude Code: `git clone https://github.com/ahkamboh/agent-caption && cd agent-caption && python setup.py` - Codex: `git clone https://github.com/ahkamboh/agent-caption && cd agent-caption && python setup.py` - Cursor: `git clone https://github.com/ahkamboh/agent-caption && cd agent-caption && python setup.py` **Models it drives:** whisper-large-v3, whisperx, mms-fa, htdemucs ##### Example output https://s3.varg.ai/how-to/previews/8298d7a02531/agent-caption.mp4 — The repo's demo video: the creator talks to camera while captions generated locally by the skill switch language line by line (English, then Spanish), with the language tag on top. First 8 s of 58.5 s. (credit: Ali Hamza Kamboh (agent-caption), https://github.com/ahkamboh/agent-caption#readme) ##### How-to tutorials that use it - [How to add animated captions to a video with AI](https://varg.ai/how-to/add-animated-captions-to-a-video.md): TikTok-style subtitles burned into an existing video: word highlight, pop words, or designed editorial captions. - [How to make a lyric video with AI](https://varg.ai/how-to/make-a-lyric-video.md): Lyrics synced to the song: the sung line big and highlighted, karaoke-style, over art or a shader. ##### More captions & subtitles skills - [3D Camera Captions](https://varg.ai/how-to/skills/hyperframes-camera-3d-captions.md): Captions that live in 3D around a talking head: the camera flies through the words, heroes pass behind the speaker. - [Remotion Captions](https://varg.ai/how-to/skills/remotion-captions.md): Transcribe speech and burn in animated captions with Remotion, or import an existing SRT file. - [ElevenLabs Speech to Text](https://varg.ai/how-to/skills/elevenlabs-speech-to-text.md): Transcribe audio and video with word timestamps, speaker labels and keyterms, ready for subtitles. - [ElevenLabs Speech to Text (inference.sh)](https://varg.ai/how-to/skills/inference-sh-elevenlabs-stt.md): Transcribe audio with ElevenLabs Scribe, speaker diarization and forced alignment, via inference.sh. - [Speech to Text](https://varg.ai/how-to/skills/inference-sh-speech-to-text.md): Transcribe audio with ElevenLabs Scribe v2 or Whisper V3 Large, with timestamps and speaker labels. - [Pretext Text Metrics](https://varg.ai/how-to/skills/pretext-yaniv-golan.md): Measure and fit text without the DOM using pretext: auto-fit sizes, know line breaks before render, canvas/WebGL text. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt ### Cuts & pacing - [Hard Cut](https://varg.ai/how-to/skills/hard-cut.md): One shot ends, the next begins on the very next frame. The default join: invisible when the content flows. One shot ends, the next begins on the very next frame. The default join: invisible when the content flows. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Cuts & pacing](https://varg.ai/how-to/browse/skills/cuts.md) - **Slug:** `hard-cut` - **Status:** live - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills/blob/main/varg-video-edit-library/references/cuts.md#hard-cut - **Works in:** Claude Code, Claude.ai, Codex, Cursor, OpenClaw, Gemini CLI, Any MCP client, varg Studio - **Requires:** ffmpeg, HyperFrames (npx hyperframes), Remotion, CapCut (by hand), Python + MoviePy - **Cost:** Free — the take renders locally; generated footage, if any, bills by the model ##### What it does Frame N is shot A, frame N+1 is shot B. No overlap, no effect. **When:** Most joins in any edit. Use it whenever the change of shot is motivated by story, speech or rhythm. **Watch out:** Two near-identical framings of the same subject cut together read as a mistake (a jump cut) unless you mean it. **Timing:** 0 frames **Also called:** straight cut, cut, butt cut ###### In FFmpeg Re-encode with the concat filter (any sources) or losslessly with the concat demuxer (identical codecs). ```bash ffmpeg -i a.mp4 -i b.mp4 -filter_complex \ "[0:v]trim=2:5,setpts=PTS-STARTPTS[v0];[0:a]atrim=2:5,asetpts=PTS-STARTPTS[a0];\ [1:v]trim=0:3,setpts=PTS-STARTPTS[v1];[1:a]atrim=0:3,asetpts=PTS-STARTPTS[a1];\ [v0][a0][v1][a1]concat=n=2:v=1:a=1[v][a]" -map "[v]" -map "[a]" out.mp4 ``` ###### In HyperFrames ```html ``` `data-start="a"` means "start when clip a ends". ###### In Remotion ```tsx import {Series} from 'remotion'; import {Video} from '@remotion/media'; ``` `trimBefore` replaced `startFrom` in v4.0.319. ###### In CapCut Put the playhead on the cut point and Split (Ctrl/Cmd+B on desktop, Split in the bottom toolbar on mobile), delete the unwanted part; clips on the main track snap together. ###### In MoviePy ```python from moviepy import VideoFileClip, concatenate_videoclips a = VideoFileClip('a.mp4').subclipped(2, 5) b = VideoFileClip('b.mp4').subclipped(0, 3) concatenate_videoclips([a, b]).write_videofile('out.mp4') ``` ###### Where this comes from - [FFmpeg filters: concat](https://ffmpeg.org/ffmpeg-filters.html#concat-filter) - [FFmpeg formats: concat demuxer](https://ffmpeg.org/ffmpeg-formats.html#concat) - [Remotion: ](https://www.remotion.dev/docs/series) - [HyperFrames: data attributes](https://hyperframes.heygen.com/concepts/data-attributes) - [CapCut Help: transitions](https://www.capcut.com/help/transitions-in-capcut) ##### Use when - Most joins in any edit. Use it whenever the change of shot is motivated by story, speech or rhythm. ##### Not for - Two near-identical framings of the same subject cut together read as a mistake (a jump cut) unless you mean it. ##### Install - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` - Codex: `npx skills add vargHQ/skills --skill varg-video-edit-library -a codex -y` - Cursor: `npx skills add vargHQ/skills --skill varg-video-edit-library -a cursor -y` ##### Example output https://s3.varg.ai/how-to/previews/5c397ed672a5/take-hard-cut.mp4 — cut between two shots, with the timeline strip of the edit. Footage: Big Buck Bunny (2008), (c) copyright 2008, Blender Foundation | www.bigbuckbunny.org, CC BY 3.0; WING IT! (2023), Blender Studio | studio.blender.org, CC BY 4.0. (credit: Rendered by varg · Big Buck Bunny (2008) + WING IT! (2023)) ##### More cuts & pacing skills - [SmartCut (capcut-ai-editor)](https://varg.ai/how-to/skills/capcut-smartcut-mcp.md): MCP that reads CapCut auto-captions to cut silences and duplicate takes right inside your CapCut project. - [Auto-Editor](https://varg.ai/how-to/skills/auto-editor.md): CLI that cuts dead space by loudness or motion and exports MP4 or editor timelines; now also an agent skill. - [AutoCut](https://varg.ai/how-to/skills/autocut.md): Edit video in a text editor: Whisper transcribes, you delete lines in Markdown, AutoCut cuts the video. - [Cut Video](https://varg.ai/how-to/skills/cut-video.md): Claude Code skill that tightens long recordings with forced-aligned cuts, keeping laughs and comedic pauses. - [Videogrep](https://varg.ai/how-to/skills/videogrep.md): Search spoken words across videos and auto-build supercuts of every match. - [PySceneDetect](https://varg.ai/how-to/skills/pyscenedetect.md): Find cuts, fades and scene changes in a video, then split it or export scene lists for the edit. - [Jump Cut](https://varg.ai/how-to/skills/jump-cut.md): Cut out a stretch inside one continuous shot, so the subject visibly jumps. Removes pauses, adds pace. Cut out a stretch inside one continuous shot, so the subject visibly jumps. Removes pauses, adds pace. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Cuts & pacing](https://varg.ai/how-to/browse/skills/cuts.md) - **Slug:** `jump-cut` - **Status:** live - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills/blob/main/varg-video-edit-library/references/cuts.md#jump-cut - **Works in:** Claude Code, Claude.ai, Codex, Cursor, OpenClaw, Gemini CLI, Any MCP client, varg Studio - **Requires:** ffmpeg, HyperFrames (npx hyperframes), Remotion, CapCut (by hand) - **Cost:** Free — the take renders locally; generated footage, if any, bills by the model ##### What it does Same camera, same framing; the subject pops to a new position or expression and time skips forward. **When:** Talking heads, vlogs, tutorials and shorts where dead air or filler words must go and speed matters more than polish. **Watch out:** Tiny jumps (a few pixels of head movement) look like glitches; either jump far enough or hide the cut with a punch-in or B-roll. **Timing:** Cut on the gap: leave 2–4 frames of silence before the next word **Also called:** jumpcut, talking-head trim, pause removal ###### In FFmpeg Keep segments [0–3.2 s] and [4.1–8 s] of one take (times from silencedetect or a transcript): ```bash ffmpeg -i take.mp4 -af silencedetect=noise=-35dB:d=0.4 -f null - 2>&1 | grep silence_ ffmpeg -i take.mp4 -filter_complex \ "[0:v]split[s1][s2];[0:a]asplit[t1][t2];\ [s1]trim=0:3.2,setpts=PTS-STARTPTS[v1];[t1]atrim=0:3.2,asetpts=PTS-STARTPTS[a1];\ [s2]trim=4.1:8,setpts=PTS-STARTPTS[v2];[t2]atrim=4.1:8,asetpts=PTS-STARTPTS[a2];\ [v1][a1][v2][a2]concat=n=2:v=1:a=1[v][a]" -map "[v]" -map "[a]" out.mp4 ``` ###### In HyperFrames Same source, two clips, independent `data-media-start`: ```html ``` ###### In Remotion Two `` of the same file with different `trimBefore` values (in frames). ```tsx ``` ###### In CapCut Split around each pause and delete it; on desktop the transcript-based editing tools can mark pauses and filler words for removal (names vary by version). ###### Where this comes from - [Wikipedia: jump cut](https://en.wikipedia.org/wiki/Jump_cut) - [FFmpeg filters: trim](https://ffmpeg.org/ffmpeg-filters.html#trim) - [HyperFrames: HTML schema](https://hyperframes.heygen.com/reference/html-schema) - [Remotion: (trimBefore)](https://www.remotion.dev/docs/offthreadvideo) ##### Use when - Talking heads, vlogs, tutorials and shorts where dead air or filler words must go and speed matters more than polish. ##### Not for - Tiny jumps (a few pixels of head movement) look like glitches; either jump far enough or hide the cut with a punch-in or B-roll. ##### Install - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` - Codex: `npx skills add vargHQ/skills --skill varg-video-edit-library -a codex -y` - Cursor: `npx skills add vargHQ/skills --skill varg-video-edit-library -a cursor -y` ##### Example output https://s3.varg.ai/how-to/previews/c4027cc6cd21/take-jump-cut.mp4 — jump between two shots, with the timeline strip of the edit. Footage: Tears of Steel (2012), (CC) Blender Foundation | mango.blender.org, CC BY 3.0. (credit: Rendered by varg · Tears of Steel (2012)) ##### More cuts & pacing skills - [SmartCut (capcut-ai-editor)](https://varg.ai/how-to/skills/capcut-smartcut-mcp.md): MCP that reads CapCut auto-captions to cut silences and duplicate takes right inside your CapCut project. - [Auto-Editor](https://varg.ai/how-to/skills/auto-editor.md): CLI that cuts dead space by loudness or motion and exports MP4 or editor timelines; now also an agent skill. - [AutoCut](https://varg.ai/how-to/skills/autocut.md): Edit video in a text editor: Whisper transcribes, you delete lines in Markdown, AutoCut cuts the video. - [Cut Video](https://varg.ai/how-to/skills/cut-video.md): Claude Code skill that tightens long recordings with forced-aligned cuts, keeping laughs and comedic pauses. - [Videogrep](https://varg.ai/how-to/skills/videogrep.md): Search spoken words across videos and auto-build supercuts of every match. - [PySceneDetect](https://varg.ai/how-to/skills/pyscenedetect.md): Find cuts, fades and scene changes in a video, then split it or export scene lists for the edit. - [Jump-Zoom Punch-In](https://varg.ai/how-to/skills/jump-zoom-punch-in.md): Hide a jump cut by scaling the next segment 110–130%. Reads as a second camera angle in talking heads. Hide a jump cut by scaling the next segment 110–130%. Reads as a second camera angle in talking heads. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Cuts & pacing](https://varg.ai/how-to/browse/skills/cuts.md) - **Slug:** `jump-zoom-punch-in` - **Status:** live - **Publisher:** HeyGen (community) - **License:** Apache-2.0 - **Source:** https://github.com/heygen-com/hyperframes/tree/main/registry/components/yt-camera-move - **Works in:** Claude Code, Claude.ai, Codex, Cursor, OpenClaw, Gemini CLI, Any MCP client, varg Studio - **Requires:** ffmpeg, HyperFrames (npx hyperframes), Remotion, CapCut (by hand) - **Cost:** Free — the take renders locally; generated footage, if any, bills by the model ##### What it does Built on [Camera Punch-In](https://github.com/heygen-com/hyperframes/tree/main/registry/components/yt-camera-move) by HeyGen (HyperFrames), Apache-2.0. The write-up and recipes below are varg's. After the cut the same shot is suddenly tighter (or wider), so the jump feels like a deliberate camera change. **When:** Every 2–3 jump cuts in a talking head, on emphasis words, punchlines or a change of point. **Watch out:** Punching past ~130% on 1080p footage shows softness; alternate in and out rather than zooming tighter every time. **Timing:** Instant (0 frames) or a 3–6 frame ease into the new scale **Also called:** punch-in, zoom cut, digital zoom cut, punch in on a jump cut ###### In FFmpeg Static 125% punch-in on the second segment (crop keeps the centre, scale restores the size): ```bash ffmpeg -i take.mp4 -filter_complex \ "[0:v]split[p][q];\ [p]trim=0:2.5,setpts=PTS-STARTPTS[w];\ [q]trim=2.5,setpts=PTS-STARTPTS,crop=iw/1.25:ih/1.25,scale=1920:1080,setsar=1[t];\ [w][t]concat=n=2:v=1:a=0[v]" -map "[v]" -map 0:a out.mp4 ``` `crop` evaluates w/h only once, so for an animated push use `scale=w='...t...':h=-2:eval=frame,crop=1920:1080` or `zoompan=...:d=1`. ###### In HyperFrames Tween the inner wrapper, never the timed clip: ```js tl.set("#seg2 .inner", { scale: 1.25 }, 2.5); // instant punch // or eased: tl.fromTo("#seg2 .inner", {scale:1.1}, {scale:1.25, duration:0.15, ease:"power2.out"}, 2.5); ``` ###### In Remotion ```tsx ``` `pushCut()` from `@remotion/transitions/push-cut` (v4.0.500+) is a ready hard cut with a short punch-in and a 2-frame flash. ###### In CapCut Select the segment after the split → Basic → Scale (e.g. 125%), or add Position/Scale keyframes for an eased push. ###### Ready-made components - [Camera Punch-In](https://skills.varg.ai/component/hf.yt-camera-move/) (hyperframes): npx hyperframes add yt-camera-move - [Push In](https://skills.varg.ai/component/hf.push-in/) (hyperframes): npx hyperframes add push-in - [Lens Zoom](https://skills.varg.ai/component/remocn.lens-zoom/) (remocn): npx shadcn@latest add @remocn/lens-zoom - [Drift](https://skills.varg.ai/component/remocn.drift/) (remocn): npx shadcn@latest add @remocn/drift ###### Where this comes from - [FFmpeg filters: crop](https://ffmpeg.org/ffmpeg-filters.html#crop) - [FFmpeg filters: zoompan](https://ffmpeg.org/ffmpeg-filters.html#zoompan) - [Remotion: pushCut](https://www.remotion.dev/docs/transitions/presentations/push-cut) - [HyperFrames: HTML schema](https://hyperframes.heygen.com/reference/html-schema) ##### Use when - Every 2–3 jump cuts in a talking head, on emphasis words, punchlines or a change of point. ##### Not for - Punching past ~130% on 1080p footage shows softness; alternate in and out rather than zooming tighter every time. ##### Install - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` - Codex: `npx skills add vargHQ/skills --skill varg-video-edit-library -a codex -y` - Cursor: `npx skills add vargHQ/skills --skill varg-video-edit-library -a cursor -y` ##### Example output https://s3.varg.ai/how-to/previews/7121432ced16/take-jump-zoom-punch-in.mp4 — punch-in between two shots, with the timeline strip of the edit. Footage: Charge (2022), Blender Studio | studio.blender.org, CC BY 4.0. (credit: Rendered by varg · Charge (2022)) ##### More cuts & pacing skills - [SmartCut (capcut-ai-editor)](https://varg.ai/how-to/skills/capcut-smartcut-mcp.md): MCP that reads CapCut auto-captions to cut silences and duplicate takes right inside your CapCut project. - [Auto-Editor](https://varg.ai/how-to/skills/auto-editor.md): CLI that cuts dead space by loudness or motion and exports MP4 or editor timelines; now also an agent skill. - [AutoCut](https://varg.ai/how-to/skills/autocut.md): Edit video in a text editor: Whisper transcribes, you delete lines in Markdown, AutoCut cuts the video. - [Cut Video](https://varg.ai/how-to/skills/cut-video.md): Claude Code skill that tightens long recordings with forced-aligned cuts, keeping laughs and comedic pauses. - [Videogrep](https://varg.ai/how-to/skills/videogrep.md): Search spoken words across videos and auto-build supercuts of every match. - [PySceneDetect](https://varg.ai/how-to/skills/pyscenedetect.md): Find cuts, fades and scene changes in a video, then split it or export scene lists for the edit. ##### Run the generation on varg varg is one API key for the models these skills drive — Seedance 2.5, Kling 4, Veo 3.1, Wan 3, Nano Banana 2.1, GPT Image 2.5, ElevenLabs v4 and 200+ more — billed in credits, with a cost estimate before anything runs. - Claude Code: `claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg` (skill + hosted MCP) - Codex, Cursor, Gemini CLI, OpenClaw: `npx skills add vargHQ/skills --skill varg-ai -a -y` - Grok and any agent that reads SKILL.md: `npx skills add vargHQ/skills --skill varg-ai`, or point it at the skill's SKILL.md - Claude.ai, ChatGPT, any MCP client: connect `https://mcp.varg.ai/mcp` - API key: https://varg.ai/dashboard/api · Model specs: https://varg.ai/models/llms.txt - [J-Cut](https://varg.ai/how-to/skills/j-cut.md): Shot B's sound starts before its picture. The ear arrives first and pulls the viewer into the next scene. Shot B's sound starts before its picture. The ear arrives first and pulls the viewer into the next scene. - **Type:** [Effect](https://varg.ai/how-to/effects.md) · [Cuts & pacing](https://varg.ai/how-to/browse/skills/cuts.md) - **Slug:** `j-cut` - **Status:** live - **Publisher:** varg (varg, verified) - **License:** MIT - **Source:** https://github.com/vargHQ/skills/blob/main/varg-video-edit-library/references/cuts.md#j-cut - **Works in:** Claude Code, Claude.ai, Codex, Cursor, OpenClaw, Gemini CLI, Any MCP client, varg Studio - **Requires:** ffmpeg, HyperFrames (npx hyperframes), Remotion, CapCut (by hand) - **Cost:** Free — the take renders locally; generated footage, if any, bills by the model ##### What it does We still see A while B's voice or ambience fades in; then the picture cuts to B. On a timeline the audio forms a J. **When:** Scene changes, interview answers (hear the reply, then see the speaker), and pulling into a location by its sound. **Watch out:** A lead longer than 1–2 s without a visual cue confuses who is speaking. **Timing:** Audio leads by 0.3–2 s; short fade-in on B's audio **Also called:** J cut, audio lead, sound advance, split edit ###### In FFmpeg Video cuts at 4 s; B's audio starts 1 s earlier. B's picture uses source 1–4 s, its audio source 0–4 s placed at 3 s: ```bash ffmpeg -i a.mp4 -i b.mp4 -filter_complex \ "[0:v]trim=0:4,setpts=PTS-STARTPTS[va];[1:v]trim=1:4,setpts=PTS-STARTPTS[vb];\ [va][vb]concat=n=2:v=1:a=0[v];\ [0:a]atrim=0:4,asetpts=PTS-STARTPTS,afade=t=out:st=3.4:d=0.6[aa];\ [1:a]atrim=0:4,asetpts=PTS-STARTPTS,afade=t=in:d=0.3,adelay=3000:all=1[ab];\ [aa][ab]amix=inputs=2:normalize=0:duration=longest[a]" -map "[v]" -map "[a]" out.mp4 ``` `normalize=0` stops amix from halving the level. ###### In HyperFrames Mute B's video and add its sound as an `