dance-video-to-prompt skill
A skill that turns a short dance or outfit-change clip into a structured prompt for AI video generators.
- 148 ★ on the repository
- Last commit
Minimal access
Why this level
- Scripts write files to `output/` and copy the skill into agent home folders.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/dance-video-to-promptDetects 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.
- Scripts write files to `output/` and copy the skill into agent home folders.
This is third-party code. Review the repository files before installing.
What it does
The `dance-video-to-prompt` skill takes a local clip (ideally under 10 seconds) and writes a prompt for generating a similar video. Scripts first extract dense frames, check sharpness, swap blurred ones for neighbouring frames and measure the audio tempo (BPM, beats). The agent then inspects the frames itself, noting body type, clothing, location, camera work and facial expression, and merges this with the rhythm. The result always has seven blocks: visual style, scene, shooting location, camera technique, an action list tied to the beat, dialogue and background sound. The repository also has an agentless variant that sends frames to an external vision model.
Who it is for. Short-video creators who rebuild similar clips in Kling, Runway or Jimeng and want the original described as a ready prompt.
Good fit when
- A dance clip must be reproduced in a video generator with beat-matched moves.
- An outfit-change or walking video is broken down into scenes.
- Prompts for several similar clips follow one seven-block template.
- You want the tempo and beat points of a finished clip before reshooting.
Not a fit when
- You need exact joint coordinates or motion capture.
- The clip runs past 20 seconds and you will not split it.
- Frames show a person whose likeness cannot be reproduced without consent.
Example request
Analyse `/path/to/dance.mp4` and return the seven prompt blocks with moves tied to the beats.Limitations
The last commit is from August 2026 and the repository declares no license. Instructions and output are in Chinese. Python 3.10+, OpenCV and Pillow are required. The skill does not render video, only prompt text. On blurry frames it avoids inventing fingers or facial features, so the description stays general. Analyse clips of real people only with your own footage or the subject's consent. The external-model mode needs your own key and endpoint, and frames are then sent to that model.
How to disable. Delete the `dance-video-to-prompt` folder from `~/.claude/skills/`, `~/.agents/skills/` and `~/.grok/skills/`. Clear `output/` working folders by hand.
FAQ
Is an API key needed?
No in skill mode, where the agent looks at frames itself. A key and model address are only needed for the `analyze_api.sh` variant.
What clip length works?
Under 10 seconds works best and up to 20 is acceptable. Cut longer videos into parts.
How do I change frame density?
By default a frame is taken every 0.33 seconds. For more precision add `--interval 0.25` to `analyze_api.sh`.
Where does the skill install?
The `install.sh` script copies it into Grok skill folders, the shared `~/.agents/skills/` and `~/.claude/skills/` when that folder exists.
What is the output?
A `prompt.md` file with seven blocks and an `analysis.json` with per-frame facts. Both sit in `output/<video name>_<time>/`.
Discussion
No questions or reviews yet.
Related
SkillsA skill that turns an idea into ready Seedance 2.0 storyboards for Jimeng: it reads references, then picks one of five methods
motion-video-production: a skill for motion videos
motion-video-production
A skill that takes an agent through seven stages for a 15 to 90 second video, a script checking each stage before the next
A skill that teaches an agent to write prompts for ByteDance's Jimeng Seedance 2.0 video model: @ references, camera language, timing, templates
A skill that turns an idea, theme or novel excerpt into a 30-second multi-shot prompt for the Seedance 2.5 video model.
A skill that turns a plot or scene into a cinematic AI video prompt for 6 to 30 seconds, with shots, performance and sound.
A skill that writes structured prompts for Seedance, Kling, Veo and Sora: a home-footage look, camera flaws, a timeline and sound
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| dance-video-to-prompt (this card) | no | 148 | |
| Awesome GPT Image | no | 2,296 | a library of ready image prompts; this skill derives a prompt from real video |
| Realistic Video Prompting | no | 337 | teaches writing video prompts by hand; this one builds them from frames and clip rhythm |
README in short
The seven output blocks are specified in `references/output_contract.md` and the template is in `templates/output_template.md`. A separate subagent, `agents/rhythm_agent.md`, handles the rhythm.