GitHub radar
Make a short AI-agent film: 59 ready styles, no video-gen or stock
Lemo-Opuscar is a toolbox of 59 code-made film styles: pick one, bring your story, and let your coding agent direct the film.
It's a set of ready film styles drawn by code (via canvas and WebGL — ways to draw graphics), not a video-generating AI. An agent writes the frames, scores the music and adds a voice.
When it helps
Useful if you make content with AI agents and want visual variety without hiring a designer or wrestling with video generation. Skip it if you don't use Claude Code or want photorealistic footage.
Pros
- 59 ready styles, from flat minimalism to full 3D animation
- No GPU needed — frames are drawn by code, not a neural net
- Any style can be remade with your own brand colors and logo
- Shows one frame, then a short demo, before you keep the film
Cons
- Quality depends on the model — styles are tuned for Opus 5.5
- One film takes an agent 30–60 minutes and real token spend
- You need Node 20+, ffmpeg and Python 3.11+ installed first
How to set it up — step by step
- 1Install Node 20+, ffmpeg and Python 3.11+ on your machine.
- 2In your terminal, run: claude plugin marketplace add lemomo-ai/lemo-opuscar, then claude plugin install lemo-opuscar@lemolab.
- 3Send the agent the text from the block below.
- 4Check the single frame and the 15–30 second demo the agent shows you before approving the final film.
- 5Keep the film if you like it, or ask the agent for another round of changes.
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Install the Lemo-Opuscar skill in Claude Code: claude plugin marketplace add lemomo-ai/lemo-opuscar, then claude plugin install lemo-opuscar@lemolab (repo: https://github.com/lemomo-ai/lemo-opuscar). Then make me a 20-30 second film in one of the styles on a topic I'll give you — show me a single rendered frame first for approval.
Light: a regular laptop with Node 20+, ffmpeg and Python 3.11+ is enough — the agent installs the rest. No GPU needed, frames are drawn by code, not by a neural model.
Open on GitHub▌ More finds