Added Oct 14, 2026
huashu-art-motion
An agent skill for making art-style and explainer animations in code, with 35 art-style recipes built in.
- Ecosystem
- Agent Skills
- Skill category
- Design
- Creator / organization
- 花叔 Huashu
What it is
huashu-art-motion is an animation Skill open-sourced by Huashu (GitHub user alchaincyf). The README says it lets a coding agent turn art styles into paintings that move. It ships 35 art-style recipe cards, from cave painting and Egyptian murals to Van Gogh, Bauhaus and 8-bit, each with a working scene and a signature transition. It also has 9 explainer-video grammars, including whiteboard, Kurzgesagt-style, 3Blue1Brown-style and finance charts; 8 of them take a JSON spec and render a frame-accurate clip in landscape or portrait. You can ask it to break down a reference animation and rebuild it in code, animate something in a particular painter's style, or lay an art-animation track over your own voiceover. When a person needs to appear on screen, the README says an image model draws the frames and the code composites them. What sets it apart is an automatic check before delivery: scripts measure stability and motion, and a film with almost no in-shot movement, one that plays like a slideshow, is blocked. The author says the Skill began as a code-only recreation of a 15-second art history animation by X user Tak, with every frame redrawn and none of the original's frames or audio included.
Who it's for:creators who want short art-style animations, explainer animations, or visuals for a voiceover video. You'll need to be comfortable with an AI coding tool that runs commands, such as Claude Code or Codex; readers who never touch a command line will find it hard going.
Before you start: install with npx skills add alchaincyf/huashu-art-motion, and set up uv, ffmpeg and Playwright Chromium before rendering. The Skill's content is in Chinese. Code and docs are MIT, but Huashu's own cartoon character, its character frames and the fan-made demo clips are for demonstration only and not covered by MIT; the README also asks you not to drop the demos' compositions, characters or assets straight into your own films. Voice and image generation are optional and need services you configure yourself, at those providers' prices. The author notes that online regression tests passed on Codex, while the Claude Code run is still waiting to be verified.
Based on the project's public documentation. Not tested by AI Inspiration World.