Huashu-motion

Huashu-motion is a motion-graphics Skill by Huashu (花叔). Your AI assistant writes animation in HTML, GSAP, Canvas, Three.js or SVG, and it is rendered frame by frame into video in the browser on your own computer, with no video generation model involved. Voice-over uses the built-in Chinese system voice on macOS and the local Kokoro model elsewhere; Whisper handles speech recognition locally, and music can use real instrument samples. In the README, the author says the 22-second demo clip was made by swapping the model behind Claude Code for DeepSeek V4.1 Flash (the 1017 build); he also writes that he mainly uses Opus 5.5.

Who it's for:Creators who want AI help making motion clips such as product intros or explainers, with precise control over every frame.

Before you start: You need Node.js 22 or later, desktop Chrome, Chromium or Edge, and FFmpeg; the instructions are in Chinese. Install with npx skills add, or git clone and create a symlink. When bin/om can't find the local program, it uses npx --yes to download and run the huashu-motion npm package. The first local voice-over downloads about 325 MB of model files and a 28 MB voice from GitHub, the Whisper model comes from Hugging Face, and instrument samples are a separate download. The author says the core scripts run on your machine and don't send your text or recordings to the author's servers; whatever you give your AI assistant still goes to the model it uses. The project is free; your assistant's subscription or API costs are separate. The code is MIT-licensed. GSAP, the fonts and the samples have their own licences, and CC-BY samples need credit. The 豆子花叔 cartoon character in the README is for demonstration only and isn't licensed for other uses.

Based on the project's public documentation. Not tested by AI Inspiration World.