Analysis updated 2026-05-18
Turn a video of a sculpture breaking apart into a site controlled by hand pinch gestures.
Build an interactive portfolio page where visitors drive an animation with palm movement.
Create a teaching demo of a mechanical part assembling that responds to mouth movement.
| senlindesign/build-gesture-animation | anil-matcha/open-poe-ai | arata-ae/purupurupngtuber | |
|---|---|---|---|
| Stars | 4 | 4 | 4 |
| Language | JavaScript | JavaScript | JavaScript |
| Last pushed | — | 2026-06-25 | — |
| Maintenance | — | Active | — |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 3/5 | 3/5 | 3/5 |
| Audience | designer | developer | general |
Figures from each repo's GitHub metadata at analysis time.
Requires Python 3.9+, FFmpeg with the libwebp encoder, and a camera-capable browser.
Build Gesture Animation is a small tool, called an Agent Skill, that plugs into AI coding assistants like Claude Code, Cursor, GitHub Copilot, and OpenAI Codex. It takes a video showing something changing continuously, like an object folding, breaking apart, or melting, and turns it into an interactive website where a viewer controls the animation with their body instead of a play button. The site reads a camera feed in the browser and tracks four kinds of input: opening and closing a palm, pinching thumb and finger together, moving two hands apart, or opening and closing the mouth. Whichever motion is chosen moves the animation forward or backward through the frames of the original video, and the direction can be reversed. The tool has a small system it says is not suited for everything. It fits things like origami, mechanical parts coming together, sculptures cracking, teaching demonstrations, and art portfolios. It does not fit long videos with narration, footage with multiple camera cuts, live physics simulations, or automatic hosting on a cloud service, since the site only runs on the user's own computer through localhost. Behind the scenes the tool checks the source video, pulls out a set number of evenly spaced frames, by default sixty but adjustable between twenty four and one hundred twenty, crops them to a consistent shape, and saves them as compressed image files. It then builds a website around those images and the chosen gesture controls, and checks that everything, including image sizes and configuration references, lines up correctly before it lets the site run. To use it, a person installs it through a command line tool called Skills CLI, or asks their AI agent to clone the repository and run its install script for whichever assistant they use. Running it requires Python 3.9 or newer and FFmpeg with its WebP encoder installed on the machine, along with a browser that supports a webcam. The project is released under the MIT license and has four stars on GitHub.
An Agent Skill that turns a video of something transforming into an interactive website you control with hand or mouth gestures via webcam.
Mainly JavaScript. The stack also includes Python, FFmpeg, JavaScript.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly designer.
This repo across BitVibe Labs
Verify against the repo before relying on details.