Vincentwei1021/mg-styles-15
15 motion design styles, each made by Claude Opus 5.5 writing code from one prompt: watch the films, copy the prompts, read the full source.
What it solves
It demonstrates how to use LLMs to generate high-quality motion design films without relying on traditional AI video or image generation models. Instead, it uses an LLM to write the actual code (SVG, Canvas, WebGL, Three.js, and Python) that renders the visuals and synthesizes the audio.
How it works
An LLM (Claude Opus 5.5) is given a detailed prompt to write a web page that can render a frame for any given timestamp. A headless Chrome instance captures these frames, and a Python script (audio.py) synthesizes the music and sound effects using instrument samples. Finally, ffmpeg muxes the frames and audio into a 10-second MP4. The process includes an AI-driven review and revision cycle using a specific rubric to refine the output.
Who it’s for
Motion designers, developers, and AI enthusiasts who want to explore programmatic animation, learn how to prompt LLMs for complex creative coding, or create custom motion graphics using code.
Highlights
- Code-based generation: Uses SVG, Canvas, WebGL, Three.js, and Blender rather than generative video models.
- Synthesized audio: Music and sound effects are generated via Python code and instrument samples.
- Complete transparency: Provides the full prompts, source code, and style notes for 15 different motion design styles.
- Iterative AI workflow: Implements a make-review-revise pipeline with an AI jury to ensure quality.
Related
- Project
- Project
- Project
- Project