---
title: "varg – AI Video Platform for Category Leaders"
description: "Bulk video generation for mobile apps and marketing. varg automatically picks the best pipeline for your team."
url: https://varg.ai
---

# Turn one idea into many.

Ads, product videos, talking heads and voiceovers from one brief. The agent plans every shot, picks the models and shows you the price before anything runs.

You see the price before anything renders.

## What’s new

- **Brief → finished ad** (Agent) – Describe the goal. The agent plans, shoots and cuts the video. [Open Agent](https://varg.ai/login?redirect=%2Fdashboard%2Fagent)
- **FLUX 3 Image Klein: small model, 4K pictures** (Image model, new) – Open weights, native 4K and edits from up to 10 references. [Try it](https://varg.ai/dashboard/tools/playground?model=flux_3_image_klein)
- **ElevenLabs v4 is now live on varg** (Voice model, new) – Direct the delivery with tags like [whispers] and [shouts]. [Try it](https://varg.ai/dashboard/tools/playground?model=eleven_v4)
- **HeyGen Video is now live on varg** (Video model, new) – AI talking avatars with natural lip-sync from a script. [Try it](https://varg.ai/dashboard/tools/playground?model=heygen_avatar)
- **GPT Image 2.5 from OpenAI** (Image model, new) – Natural light, legible type and edits that touch only what you ask. [Try it](https://varg.ai/dashboard/tools/playground?model=gpt_image_2_5)
- **Seedance 2.5: long shots, native audio** (Video model) – 4–30 s shots up to 1080p, with native audio. [Try it](https://varg.ai/dashboard/tools/playground?model=seedance_2_5)
- **varg inside Claude** (MCP) – Video, images and voice without leaving your chat. [Connect](https://varg.ai/dashboard/mcp)
- **FLUX 3: video with sound** (Video model, new) – Text, image or video in – synced audio out. [Try it](https://varg.ai/dashboard/tools/playground?model=flux_3)
- **Cast a character that stays on-model** (Presets, new) – 29 characters that look the same in every shot. [Browse presets](https://varg.ai/login?redirect=%2Fdashboard%2Fresources%2Fpresets)
- **39 templates, all made with varg** (Templates) – Watch one, then make your version of it.
- **Plans from $24 a month** (Pricing) – 700 credits a month, one balance everywhere. [See pricing](https://varg.ai/dashboard/pricing)

## Try an example

- 5-second product reveal for a sparkling drink – ≈218 credits · Seedance 2.5 · Nano Banana 2
- UGC talking head for a skincare launch – ≈176 credits · VEED Fabric · ElevenLabs v3
- Voiceover ad for a fitness app challenge – ≈356 credits · Kling v3 · ElevenLabs v3
- 30-second explainer: what caffeine does to sleep – ≈1,239 credits · Seedance 2.5 · ElevenLabs v3
- Wildlife-doc spot with my product as the hero – ≈66 credits · MiniMax H3 Max · Nano Banana 2

## Find an example for your task

The latest work made with varg. Open one to watch it, then start your own version from it – same format, your product.

[All 39 templates](https://varg.ai/login?redirect=%2Fdashboard%2Fcommunity)

- [High-Speed Splash – Freeze + Reverse](https://s3.varg.ai/files/db4ef430-d0c7-403e-a8e2-722292a92e8c/file_f600204d5e7c.mp4) (Product · 0:05) – A can plunges into a glass tank of water. Extreme slow motion: the crown splash blooms upward, then time reverses and the can rises out dry. 1/16000s look, back-lit strobes, pure black background. Product card locks the can. Demonstrates high-speed photography, time reversal, and native audio.
- [Balloon Lift – Cliff Above the Clouds](https://s3.varg.ai/files/db4ef430-d0c7-403e-a8e2-722292a92e8c/file_3abbdfd2da52.mp4) (Product · 0:07) – A vast sea of clouds at dawn. A can stands on a stone ledge over the void with balloons tied to it, and lifts off slowly, drifting over the cloud sea. Locked camera, cold blue-grey palette. Product card locks the can. Demonstrates landscape-scale product placement, atmospheric perspective, and native audio ambience.
- [Impossible Cut – Can Sliced Like a Fruit](https://s3.varg.ai/files/db4ef430-d0c7-403e-a8e2-722292a92e8c/file_3f01288ac57f.mp4) (Product · 0:05) – Macro food ASMR on a dark slate board. A chef’s knife presses down through the aluminum can in one clean vertical cut, and the two halves fall open to reveal a ripe orange citrus inside. Product card locks the can’s exact shape and finish. Demonstrates macro photography, locked-camera ASMR, and surreal product reveal.
- [Ostrich – One Gulp](https://s3.varg.ai/files/db4ef430-d0c7-403e-a8e2-722292a92e8c/file_0ca920d85a22.mp4) (Product · 0:05) – Tight studio portrait of an ostrich holding the can in its beak. It throws its head back and chugs the whole can in one continuous swallow, a visible bulge travelling down the neck. Identity locked with character + product cards. Demonstrates locked-camera studio lighting, fast head motion, and native audio.
- [Gorilla Vlog – Shibuya Fisheye](https://s3.varg.ai/files/db4ef430-d0c7-403e-a8e2-722292a92e8c/file_6d483797d9c9.mp4) (Product · 0:06) – Handheld fisheye vlog at Shibuya crossing at dusk. The pink gorilla holds the product can up to the lens and ad-libs a line. Identity locked via a character card + product card: the still is generated with both references, then the video gets the still as its first frame plus the same cards with explicit @imageN roles. Demonstrates omni_reference multi-image identity lock, fisheye lens distortion, and native audio dialogue.
- [Glossy Photoshoot – Character Cast](https://uu.varg.ai/templates/footer-photoshoot/series-1080p.mp4) (Memes · 0:34) – Seven-clip photoshoot where a cast of 29 characters does a high-fashion magazine shoot on a mirror-black set. Every clip starts and ends on the same white FLASH frame, so clips concatenate in any order and the strobe itself is the transition. Each clip packs 2-3 character groups with a comedy gag per group. Demonstrates multishot Seedance with strobe-cut transitions, character identity anchoring via neutral cards, and 1080p rendering.
- [varg – Hero Background Loop](https://uu.varg.ai/templates/footer-loop/hypercars.mp4) (Cinematic · 0:26) – Seamless 26-second loop built from three Seedance 2.5 beats chained by pixel-matched keyframes: a hypercar column pulls back to a wheel macro, a creator conveyor crane-ups to a pupil, and a style-morph cascade flows through five rendering styles back into the opening glow. Every keyframe is generated in the template, so the whole loop is reproducible. Demonstrates first/last-frame chaining for seamless loops, omni_reference multi-image prompts, and keyframe-driven scene transitions.
- [Calisthenics Gym – Blue Silhouette Workout](https://uu.varg.ai/templates/calisthenics-gym/calisthenics-gym.mp4) (Cinematic · 0:35) – Seven calisthenics exercises animated as flat blue silhouettes against a golden-hour outdoor workout park. Each exercise is generated as a nano-banana-2 image (pose → blue silhouette edit), then animated with MiniMax H3 Max image-to-video. The template chains 7 five-second clips with fade transitions over an energetic workout beat. Demonstrates the image-edit-to-silhouette pipeline, H3 Max motion prompts for exercise animation, and multi-clip composition with music.
- [L-theanine – Cartoon Explainer](https://uu.varg.ai/templates/l-theanine/skeleton-explainer.mp4) (Cinematic · 0:30) – Audio-driven clinical 3D medical-visualisation across three seedance takes. Beat timestamps are computed from each speech segment duration rather than authored, so picture cuts land where the sentence lands and the clip is trimmed back to the read. Demonstrates computed beat timing, seedance 5/10/15s rounding, and music ducking over a near-continuous VO.
- [Brainrot Health Explainer – Sloop](https://uu.varg.ai/templates/brogevity/sloop-melatonin.mp4) (Memes · 0:20) – Italian-brainrot health explainer with one consistent AI character: a nano-banana-pro reference image drives every minimax h3-max scene, and the narrator audio is generated natively by the video model from lines written directly in the prompts – no TTS, no captions pipeline. Scenes share one setting with time-of-day progression for continuity.
- [Omega-3: Jesus vs The Devil – AI Podcast](https://uu.varg.ai/templates/omega3-jesus-devil/omega3-podcast.mp4) (Two-Person Dialogue · 14:59) – A 15-minute AI-generated educational podcast where Jesus asks questions about Omega-3 and The Devil gives misleading answers using real studies out of context. 78 scenes generated with MiniMax H3 Max (native speech synthesis – no separate TTS). Each scene uses a character reference image and embeds dialogue directly in the video prompt. Demonstrates long-form podcast generation, character consistency across 78 clips, per-phrase duration matching (~1.8 words/sec), and the deceptive-claim-then-truth format for educational content.
- [Action Jesus vs Devil Seedance 2.5](https://s3.varg.ai/files/194b94d0-58ff-4e8b-91de-4f1959d04013/file_a1cc15626063.mp4) (0:10) – Hook video for fitness app
- [Native Audio Captions with Prompt Correction](https://uu.varg.ai/templates/flo-captions.mp4) (UGC Ads · 0:08) – Source video with native dialogue, auto-transcribed via Whisper with a prompt hint that corrects a misheard name (Flo vs flow). Demonstrates vid.audio lazy-derived captions + the prompt prop for transcription accuracy.
- [Morning Stiffness – Voiceover Hook](https://uu.varg.ai/templates/tai-chi/ep4.mp4) (UGC Ads · 0:17) – Misty mountain-porch scene with an ElevenLabs voiceover over muted b-roll, plus one native-audio clip. Shows mixing generated VO with in-clip dialogue in one timeline.
- [The Bet – Before & After](https://uu.varg.ai/templates/tai-chi/ep3.mp4) (Before & After · 0:33) – Mirror-selfie challenge that hard-cuts to day 28 in identical framing. The after-shot is chained off the before-composite so the pose and background match exactly.
- [AI Character Showcase Reel](https://s3.varg.ai/renders/1775534627355_character_showcase.mp4) (0:32) – Multi-scene AI persona showcase with animated clips, voice range demos, and talking segments
- [Claymation UGC Ad – Super Focus](https://su.varg.ai/renders/1775274181981_idqe0r1a.mp4) (0:17)
- [Nature Doc – Waterhole (H3 Max)](https://s3.varg.ai/files/db4ef430-d0c7-403e-a8e2-722292a92e8c/file_501c6c37a8b0.mp4) (Cinematic · 0:07) – H3 Max image-to-video original. Wildlife documentary: a gorilla drinks from a waterhole, spots an ostrich, then picks up an orange can and drinks from that instead. 600mm telephoto, heat haze. 16:9, 6s, native audio.
- [Gorilla Vlog – Shibuya Fisheye (H3 Max)](https://s3.varg.ai/files/db4ef430-d0c7-403e-a8e2-722292a92e8c/file_419be31ec594.mp4) (Cinematic · 0:07) – H3 Max image-to-video original. Handheld fisheye vlog at Shibuya crossing: the gorilla talks to camera about its drink, takes a sip, keeps walking. Dense crowd ambience. 16:9, 6s, native audio.
- [Tilt-Shift Diorama – Miniature City](https://s3.varg.ai/files/db4ef430-d0c7-403e-a8e2-722292a92e8c/file_bc8c64ddbaf8.mp4) (Cinematic · 0:07) – H3 Max image-to-video original. Tilt-shift miniature city from above: tiny cars, a gorilla, ghost and donut visit a giant can as tourists. Stop-motion toy cadence, warm golden light. 16:9, 6s, native audio.
- [Giant Can – Downtown Scale](https://s3.varg.ai/files/db4ef430-d0c7-403e-a8e2-722292a92e8c/file_49db6d24edcc.mp4) (Cinematic · 0:07) – H3 Max image-to-video original. A giant beverage can towers over downtown towers; slow tilt up, condensation drop, splash at the base where a sandwich character looks up. 16:9, 6s, native audio.
- [Portal – Camera Follows Her into the Brand World](https://s3.varg.ai/files/db4ef430-d0c7-403e-a8e2-722292a92e8c/file_262de6866cc5.mp4) (Product · 0:05) – Steadicam follows a woman from behind across a grey rooftop through a tall oval portal into a sun-drenched orange desert canyon. Anamorphic 40mm, horizontal lens flare crossing the rim. Identity locked with character + product cards. Demonstrates follow-cam, portal transition, and dual-light colour grading.
- [Billboard Break – Jimmy Climbs Out](https://s3.varg.ai/files/db4ef430-d0c7-403e-a8e2-722292a92e8c/file_10a0ad5dd2ad.mp4) (Product · 0:05) – A huge illuminated billboard shows the gorilla, but it breaks the frame: it heaves itself out into real space, climbs down the building, and lands on the sidewalk in front of pedestrians. Cinestill 800T night look. Identity locked with character + product cards. Demonstrates cinematic narrative, blue-hour grading, and native audio dialogue.
- [Fridge Cam – POV from the Shelf](https://s3.varg.ai/files/db4ef430-d0c7-403e-a8e2-722292a92e8c/file_b89227f2d679.mp4) (Product · 0:06) – Wide-angle camera mounted inside a refrigerator looking out. The gorilla opens the door, reaches past the lens for the can, and swings the door shut – frame goes black. Identity locked with character + product cards. Demonstrates POV camera placement, cold/warm lighting split, and native audio SFX.

## See how it’s made

Every run is a plan you can read: what it is built from, the prompt the agent wrote from a one-line brief, each step with its model and price, and the finished video.

### Product splash

**Materials:** Product card, First frame
- Liquid: Clear water
- Headline: Cold. Fast. Yours.

**Brief → prompt**

- You write: 5-second product reveal for a sparkling drink
- The agent writes: A can plunges into a glass tank of water. Extreme slow motion: the crown splash blooms upward, then time reverses and the can rises out dry. Back-lit strobes, pure black background.

**Shots and steps** (5 steps)

- Product card – reused from the sample: included
- First-frame still – Nano Banana 2: 9 credits
- 5 s shot – Seedance 2.5 · 720p: 203 credits
- Splash sound – native audio, included with the shot: included
- Assemble and encode – varg render: 6 credits

This run: ≈ 218 credits

**Finished video:** 0:05 · 9:16 – [Watch with sound](https://s3.varg.ai/files/db4ef430-d0c7-403e-a8e2-722292a92e8c/file_f600204d5e7c.mp4)

You can change: Your product, from photos; The liquid; Headline and end card; Premium 9:16 or fast 16:9.
Stays as tested: Camera move and timing; Black studio set; Length: 5–6 seconds.

[Open this example and change it](https://varg.ai/login?redirect=%2Fdashboard%2Fagent)

### Talking head

**Materials:** Presenter, Logo
- Script: “Varg AI is just insane…”
- Voice: Warm and conversational

**Brief → prompt**

- You write: Selfie-style talking head that sells our app, with captions
- The agent writes: UGC selfie video: a friendly presenter tells viewers why they should try the product, with bold captions and light background music.

**Shots and steps** (6 steps)

- Presenter – reused from the sample: included
- Voiceover – ElevenLabs v3: 3 credits
- Lip-synced video – VEED Fabric 1.0: 158 credits
- Music bed – ElevenLabs Music: 9 credits
- Captions – timed from the voiceover words: included
- Assemble and encode – varg render: 6 credits

This run: ≈ 176 credits

**Finished video:** 0:11 · 9:16 – [Watch with sound](https://s3.varg.ai/renders/1775534627355_talking_head_promo.mp4)

You can change: Script; Presenter and voice; Logo and end card.
Stays as tested: Selfie framing; Caption style; One continuous take.

[Open this example and change it](https://varg.ai/login?redirect=%2Fdashboard%2Fagent)

### App ad with voiceover

**Materials:** Porch, wide, To camera
- Voiceover: “If you wake up feeling too stiff to move…”
- Line to camera: “It feels so good.”

**Brief → prompt**

- You write: Voiceover ad for a fitness app challenge
- The agent writes: Open on a misty porch at sunrise with two people doing tai chi. The voiceover hooks people who wake up stiff, one of them says “It feels so good” to camera, then the ad invites viewers to the challenge.

**Shots and steps** (7 steps)

- 2 scene images – Nano Banana Pro: 30 credits
- Wide shot, 12 s – Kling v3: 213 credits
- Line to camera, 5 s – Kling v3 with voice: 89 credits
- Voiceover – ElevenLabs v3: 3 credits
- Music bed – ElevenLabs Music: 9 credits
- Captions for the spoken line – Whisper: 6 credits
- Assemble and encode – varg render: 6 credits

This run: ≈ 356 credits

**Finished video:** 0:17 · 9:16 – [Watch with sound](https://uu.varg.ai/templates/tai-chi/ep4.mp4)

You can change: What the app does and who it is for; Voiceover and spoken line; End card.
Stays as tested: Two-scene structure; Vertical 9:16; About 17 seconds.

[Open this example and change it](https://varg.ai/login?redirect=%2Fdashboard%2Fagent)

## Use varg inside Claude

Connect Claude, ChatGPT, Cursor or VS Code and generate video, images and voice where you already work. One sign-in, no API key, the same credits as the app.

Connector URL: `https://mcp.varg.ai/mcp`

- [Add connector in Claude](https://claude.ai/customize/connectors?modal=add-custom-connector&connectorName=varg&connectorUrl=https%3A%2F%2Fmcp.varg.ai%2Fmcp)
- [Setup for every client](https://varg.ai/mcp)

## Plans from $24 a month.

You approve the price before anything renders. [See plans](https://varg.ai/dashboard/pricing)

## Questions before the first run

### Do I need an account to try it?

Not to look around: write a brief, open any template and see how it was made. You sign in to run it, and the brief you wrote comes with you to the agent.

### How much does a video cost?

It depends on the shots and models, and the agent shows the total before anything renders. A 5-second clip runs from about 37 credits to about 200 on premium models; a finished 30-second ad is usually around 500.

### Can I use my own product, logo and voice?

Yes. Attach product photos, screenshots or a logo to the brief, and pick a presenter or voice from Presets or describe a new one. Everything you upload stays in your workspace.

### Is there a watermark? Can I use the videos commercially?

No watermark, full-quality export, and commercial use on every plan – Starter included.

### What happens if a render fails?

Credits are held when a render starts and only charged when it completes, so a failed render costs nothing. You change the brief or a single shot and run just that part again.

### Which models does it use?

More than 100 – Seedance, Kling, Veo, FLUX, MiniMax, Nano Banana, GPT Image, ElevenLabs and more. The agent picks one per shot; you can switch any step to another model and see the price change first.

### Can I use varg from Claude, the API or my own code?

Yes. The same credits work in the app, over the API and SDK, and inside Claude, ChatGPT, Cursor or VS Code through the varg MCP server.

## One brief. Every format.

Describe the ad or start from a template. You see the price before anything renders.

[Write a brief](https://varg.ai/login?redirect=%2Fdashboard%2Fagent) · [book a 15-minute demo](https://cal.com/alex-varga/15)

Backed by people from: Snapchat, together.ai, Spotify, E2B, Lensa AI.