Seedance.skill
A skill that drives AI video through a clay-render scene: intake, geometry checks before render and a prompt from the same data.
- 62 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The skill runs local Node scripts for scene checks and calls no external services.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/seedance-skillDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill runs local Node scripts for scene checks and calls no external services.
This is third-party code. Review the repository files before installing.
What it does
Seedance.skill helps steer AI video through a reference clip, a clay render: it fixes only spatial relations, camera path and rhythm while materials, light and color come from reference images. The skill tackles three typical problems. Requirements are clarified with three questions: length, platform and a single shot or cuts. Scene errors are caught by `engine/probe.mjs` before rendering: a subject leaving frame, too small, or a camera going into a wall. The prompt is assembled by `engine/make_prompt.mjs` from measured geometry, so it promises no moves absent from the clay render. There are four lines of work: the prompt, a clay render from `scene.json`, reference images and rebuilding a clay render from a single photo. There are ten camera types, all defined as functions of time in `engine/solver.mjs`.
Who it is for. Video makers and ad developers who steer the camera in Seedance through a reference clip and lose time to rework.
Good fit when
- A clip needs an exact orbit around a hero with a set length and rhythm.
- A clay render should be checked for geometry errors before an expensive render.
- The prompt must match what the camera actually does in the clay render.
- A photo should become a scene with camera movement.
Not a fit when
- A quick prompt with no clay render is wanted: simpler to write by hand.
- Materials and light must be set by text alone.
- There is no Node 18 or ffmpeg for frame checks.
Example request
I want to shoot a 10-second orbit of a hero with no cuts. Clarify details, build the clay render, check it before rendering and write the prompt.Limitations
The skill does not make video itself: it prepares the clay render and prompts and Seedance makes the clip. Node 18 or newer and ffmpeg for frame checks are needed, plus Python 3 with opencv-python for photo-based rebuilding. `npx hyperframes` downloads automatically. The clay render pins position and timing but cannot hold a pose against common sense: the author cites a case where the model quietly turned a figure around. The clay render color bleeds into the result. Documentation is bilingual, Chinese and English.
How to disable. Delete the `~/.claude/skills/seedance-skill` folder and restart the agent. Created scenes and clips remain in your projects.
FAQ
How do I install it?
Run `git clone https://github.com/alchaincyf/seedance-skill.git ~/.claude/skills/seedance-skill` and restart the agent.
What is a clay render?
A reference clip made of simple shapes that fixes space, camera path and rhythm while the look comes from references.
What does `engine/probe.mjs` check?
A subject leaving frame, too small a size, dead time, a camera inside a wall, a formation over the frame and mismatched aspect ratios.
What are the four lines of work?
The prompt, a clay render, reference images and rebuilding a clay render from a photo.
Which agents does it work with?
Claude Code, Codex, Cursor, OpenClaw and others that read the Agent Skills format.
Discussion
No questions or reviews yet.
Related
SkillsA skill that improves Seedance prompts, plans storyboards and generates video through Atlas Cloud, plus a library of ready prompts
A skill that turns a video idea into Seedance 2.5 settings, reference roles and one paste-ready prompt for the video model.
A skill that teaches an agent to write prompts for ByteDance's Jimeng Seedance 2.0 video model: @ references, camera language, timing, templates
A Codex skill that turns an idea into a Seedance 2.5 prompt with a 3D interface, a CRT monitor and a continuous shot plan.
A Codex skill that plans fight rhythm, camera and effects in Seedance 2.5 prompts and revises the prompts you already have.
A dialogue that turns an idea into a storyboard and a prompt for the Seedance 2.0 video model in Jimeng, with six templates and ten examples
README in short
The README covers why a clay render helps, three problems and their fixes, four lines of work, a quick start, installation, dependencies, facts learned in practice and the folder layout.