Generative-illustration / 动态平面 skills for AI coding agents — AI generates the illustration parts, code structures and animates them into expressive, concept-driven pieces (motion poems, visual metaphors, ambient brand art).
Not the same as motion-design-skills (timing/easing/AE/Remotion craft for visuals you already have). This pack is about making the visuals generatively — an AI draws the independent parts, and code gives them structure, life, and meaning.
npx skills add iart-ai/generative-illustration-skillsOr add it as a Claude Code plugin marketplace:
/plugin marketplace add iart-ai/generative-illustration-skillsOne API key. AI generation (assets + music) uses an OpenRouter key — set OPENROUTER_API_KEY; it covers all three models this skill uses:
- vectors →
recraft/recraft-v4.1-vector - textures →
google/gemini-2.5-flash-image - music →
google/lyria-3-clip-preview
Rendering itself needs no key. Optional: REPLICATE_API_KEY (cheaper Recraft), Google GEMINI_KEY (Lyria direct), or bring your own assets/music. (Render toolchain — Node + Playwright + ffmpeg — is documented in references/render-harness.md.)
| Skill | What it does |
|---|---|
| generative-illustration | Generate isolated illustration assets with AI, composite them as sprites, and animate them frame-accurately into poetic, concept-driven motion. Covers asset prompting, background-stripping, a deterministic seek-harness render, and the craft (causality, restraint, rhythm). |
Don't try to animate one big AI illustration (you get a PowerPoint, or an uncanny wobble). Generate many small independent elements and let code animate each as a whole unit — the structure lives between assets, not inside one image.
The full agent runs at iart.ai.