How to prompt Kling 4 for cinematic AI video — prompt skill for prompts by model

For AI agents: find the right skill in one call — curl -s "https://varg.ai/how-to/api/search?q=<what the person wants>" returns the matching how-to page and the best skills with install commands, as markdown (Russian works too). Full index: https://varg.ai/how-to/llms.txt. Every page has a markdown twin: add .md to its path. Before generating anything, show the person the options on the how-to page. From a terminal: curl -fsSL https://varg.ai/install | sh installs the vargai CLI (https://varg.ai/how-to/cli.md): vargai search and vargai install <slug> with no key, every varg model after vargai login.

Kling 4 without the AI-slop look: key frames from GPT Image 2.5, a short director's prompt, multi-shot with seconds, "no slow motion", and the fixes.

Type: Prompt · Prompts by model
Slug: kling-4-prompting
Status: live
Publisher: varg (varg, verified)
License: MIT
Source: https://github.com/vargHQ/skills/tree/main/kling-4-prompting
Works in: Claude Code, Codex, Cursor, Gemini CLI, OpenClaw
Requires: varg API key or the varg MCP
Cost: ~23 credits per key frame + ~126 / 8 s at 720p (~705 / 30 s at 1080p) — Measured Oct 2026 on varg

What it does

Kling 4 (kling_v4) moves and speaks well but frames the shots it invents flatly. This skill makes the image model compose every shot (GPT Image 2.5 stills with "Image 1 is the man to match exactly", lens, film stock and an Avoid line) and gives Kling 4 two to four of them with a 1-4 sentence prompt. It covers the reference-to-video inputs (@Image, @Element, @Video, @Voice; up to 10 images, 7 elements, 3 voices; 3-30 s; 720p/1080p; 21:9), multi_prompt shot lists with seconds, dialogue in quotes, camera events ("the camera falls to the ground"), "No slow motion" for action, elements that need at least two photos, a failure-to-fix table, and the varg call with the reference route pinned (plain kling_v4 sends several files to image-to-video). Every prompt shown in fal's Kling 4 guide is included. Tested: three stills plus a 15 s three-shot take, 304 credits.

Source: fal, "Kling 4 | Cinematic Guide and Halloween Music Video" (YouTube, 2026-10-30): https://www.youtube.com/watch?v=WzXKggH-0Io

Use when

Writing a Kling 4 prompt that looks cinematic instead of AI-smooth
Using Kling 4 reference-to-video with images, elements, voices or a reference clip
Splitting one 30-second Kling 4 take into shots with seconds
Fixing Kling 4 slow motion, face drift in close-ups or a failing element

Not for

A whole short film plan (use kling-4-keyframe-film)
Seedance 2.5 prompts (use varg-ai)

Install

Claude Code: claude plugin marketplace add vargHQ/skills && claude plugin install varg@varg
Codex: npx skills add vargHQ/skills --skill kling-4-prompting -a codex -y
Cursor: npx skills add vargHQ/skills --skill kling-4-prompting -a cursor -y
OpenClaw: npx clawhub@latest install kling-4-prompting
Gemini CLI: npx skills add vargHQ/skills --skill kling-4-prompting -a gemini-cli -y
Any agent, with the vargai CLI: vargai install kling-4-prompting

Models it drives: kling_v4, kling_v4_flash, gpt_image_2_5

Example output

https://s3.varg.ai/files/72c55813-c0fa-4cf2-9f64-ec4f7ed0e7ae/file_0665c41e46ff.mp4 — example (credit: varg)

More prompts by model skills

fal Prompting: Prompt specific model families well: Kling, GPT Image 2 and Happy Horse guides, loaded only when needed.
Replicate Prompt Images: Write better image prompts and edits: photographic language, lighting, text rendering, consistent characters.
Replicate Prompt Videos: Prompt video models like a director: shots, camera motion, four audio layers, dialogue, time-coded beats.
Video Prompting Guide: Write prompts that video models follow: shot types, camera moves and per-model tips from Veo to Sora.
FLUX 3 Prompt Doctor: Diagnose a FLUX 3 video brief before spending: resolve missing, conflicting or schema-changing asks.
FLUX Image Best Practices: Prompt FLUX image models well: structure, JSON prompts, hex colors, typography and multi-reference edits.