Skip to content

Back to the catalog

lemomo-ai/lemo-opuscar
github.comrepo

lemomo-ai/lemo-opuscar

test

TL;DR test
Document

What it claims

lemo-opuscar is a collection of 39 film styles, each shipped as a reusable style prompt (STYLE.md) plus a short demo film made entirely in code by Claude Opus 5.5. You pick a style, supply your own story, and a coding agent "directs" the film: Canvas and WebGL pages rendered frame by frame, original scores written from free sample libraries, and local text-to-speech for narration. No video-generation models and no stock footage are used. The styles are explicitly tuned for Opus 5.5 โ€” the author states other models may not reproduce them.

How it works

Given a spoken request (e.g. "make a 45-second film in the watercolor style about the coffee farm my family runs, warm female narrator"), the agent asks one round of up-front questions โ€” undecided topic details, whether you're bringing your own voice/music/material, and whether you want a storyboard first โ€” then reads three guides and works "like a small studio" through drawing, scoring, mixing and rendering. If you ask for a storyboard it stops once to show key shots in the real style; otherwise it goes straight to the finished video. Styles are named in English or Chinese and span categories including hand-drawn/painting, East Asian traditions, printmaking, graphic/type, keynote, cartoon/anime, games, cinema eras, and materials/3D.

How to use it

Two entry paths. The recommended one is the skill:

claude plugin marketplace add lemomo-ai/lemo-opuscar
claude plugin install lemo-opuscar@lemolab

First use downloads guides, tools and style prompts (~60 MB) to ~/lemo-opuscar, shared across films; each film's project lives in the folder you started from. For non-Claude agents, copy plugin/skills/lemo-opuscar/ into their skills folder. Alternatively, clone and run in-repo:

git clone https://github.com/lemomo-ai/lemo-opuscar.git
cd lemo-opuscar
claude

Films then land in films/<name>/. Requirements: Node 20+, ffmpeg, Python 3.11+; the agent installs packages and the headless browser it renders with. Voice model and sample libraries download only when a step needs them.

Limits

  • One film takes an agent roughly 30โ€“60 minutes and a "fair amount"/"not small" token usage โ€” plan cost accordingly.
  • Style fidelity is only claimed for Opus 5.5; other models are not guaranteed to match.
  • Licensing is split: code is MIT; guides, STYLE.md files and films are CC BY 4.0. Third-party samples, fonts and music keep their own licences (CC0, CC BY, OFL), listed per-demo in CREDITS โ€” check these before reusing assets commercially.

Organisations named here

  • LemoLab

Technologies named here

Techniques named here

  • Frame-by-frame code-rendered animation
  • Local text-to-speech narration
  • Upfront clarification with optional storyboard preview