lemomo-ai/lemo-opuscar

Lemo‑Opuscar – AI‑driven code‑generated short films

What it is

• A collection of 39 “film styles”, each accompanied by a short demo film that was entirely created by an AI coding agent (Claude Opus 5.5). The agent writes the JavaScript/HTML code that renders each frame with Canvas or WebGL, composes original music from free sample libraries, and adds a locally‑generated TTS voice‑over. No stock footage or pre‑made video is used.

How it works

• The repository ships three guide files (DIRECTOR.md, TECHNIQUE.md, and a per‑style STYLE.md) that tell a Claude‑style agent how to:
  1. Direct – decide story, pacing, camera moves, narration, and music.
  2. Build – render frame‑by‑frame graphics, synthesize voice, mix audio, and export a video with ffmpeg.
  3. Style – apply the visual and auditory characteristics of a chosen style (e.g., Watercolor, Ukiyo‑e, 16‑bit Pixel RPG, etc.).
• Users simply give a natural‑language prompt like “Make a 45‑second film in the watercolor style about my family’s coffee farm, warm female narrator.” The agent reads the guides, asks a single clarification question, optionally shows a storyboard, then runs the whole pipeline.

Installation / usage

• Option 1 (recommended): Install as a Claude plugin/skill:

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

  The first run downloads 60 MB of guides, tools, and style prompts to `/lemo-opuscar`. Subsequent films are created in the folder where you invoke the skill.
• Option 2: Clone the repo and run claude from the repository root. Films are stored under films/<name>/.
• After installation, just speak or type your film request; the agent handles the rest.

Technical requirements

• Node 20+ (for the rendering scripts), Python 3.11+ (for auxiliary tools), and ffmpeg (for video encoding).
• The agent automatically installs needed npm packages and a head‑less browser for rendering.
• Large assets (voice model, sample‑library audio) are downloaded on demand.

What you get

• A complete project folder for each film: source code, intermediate frames, audio tracks, and the final video file.
• A visual gallery of all 39 styles at https://lemomo‑ai.github.io/lemo‑opuscar/.

Licensing

• Code is released under the MIT license.
• Guides, style descriptions, and the generated demo films are CC BY 4.0.
• Any third‑party samples, fonts, or music retain their original licenses (CC0, CC BY, OFL) and are listed in each demo’s CREDITS file.

Who it’s for

• Creators who want to prototype short animated stories without learning graphics programming.
• Researchers exploring AI‑driven content pipelines, especially code‑first generation rather than pixel‑based diffusion.
• Anyone curious about how a large‑language model can act as a mini‑studio, handling direction, art, sound, and editing automatically.