43 Film Styles, Each With a Short Film Made Entirely in Code
You've probably seen AI video tools that generate clips from a text prompt. But what if the film wasn't generated at all—what if it was written, frame by frame, as code? That's the premise behind Lemo-Opuscar, a project that pairs 43 distinct film styles with short films made entirely through code written by an AI agent.
What It Does
Lemo-Opuscar is a collection of 43 film styles, each accompanied by a short film created entirely in code. The styles are tuned for Claude Opus 5.5, and every film was directed, drawn, scored, and mixed by an AI agent writing code—no video generation, no stock footage involved.
The technical approach is straightforward but unusual: films are rendered as canvas and WebGL pages, frame by frame. Music comes from free sample libraries, and narration uses text-to-speech. The project includes a featured film called OPUSCAR 98, a 6-minute 25-second piece where a character named Clawd walks through all 98 Best Picture winners from 1927 to 2025, with each film redrawn in a style that fits the original movie.
The project is bilingual (English and Chinese) and designed to work with coding agents. You pick a style, bring your own story, and let your agent direct the film.
Why It's Cool
It's not a curated list—it's a working pipeline. The README is explicit about this: "Not an awesome list. Every film here was made by me, with Claude Opus 5.5." This matters because it means the styles aren't just descriptions or references. They're prompts and tools that have actually been used to produce finished films. You're not browsing inspiration; you're getting a system that's been tested end-to-end.
The style variety is genuinely broad. The project recently added Copperplate Engraving, Sci-fi Hologram HUD, Mid-century Cartoon, and Silkscreen Travel Poster styles. That range—from historical print techniques to retro-futuristic interfaces—suggests the agent can handle very different visual languages. The OPUSCAR 98 film demonstrates this by matching each Best Picture winner with a fitting style, which is a clever way to showcase what the system can do.
The workflow respects your time. When you start a film, the agent asks you once, up front, about anything it can't decide—your topic details, whether you have your own voice or music assets, and whether you want to see a storyboard first. If you say yes to the storyboard, it stops once to show you key shots in the real style. Otherwise, it goes straight to the finished film. That's a thoughtful balance between control and automation.
It's designed for coding agents, not just humans. The recommended installation is as a Claude plugin, which means you install it once and then use it from any folder. The first use downloads guides, tools, and style prompts (about 30 MB) to a shared location. Each film's project stays in the folder where you started. For other agents, you can copy the skill folder into their skills directory. This is infrastructure for agent-assisted creative work, not just a demo.
The constraint is the point. No video generation models, no stock footage—everything is code. That means every frame, every note, and every cut is reproducible and inspectable. You can see exactly how the film was made because it's all in the source. That's a fundamentally different relationship with the output than you get from a black-box video generator.
How to Try It
There are two ways to get started. The skill installation is recommended because it's the least friction.
Option 1: Install as a skill (recommended)
In your terminal:
claude plugin marketplace add lemomo-ai/lemo-opuscar
claude plugin install lemo-opuscar@lemolab
After that, you can use it from any folder. On first use, it downloads the guides, tools, and style prompts (about 30 MB) to ~/lemo-opuscar, shared across all your films. Each film's project—from source to finished video—goes in the folder you started from.
For other agents, copy plugin/skills/lemo-opuscar/ into their skills folder.
Option 2: Clone the repo
git clone https://github.com/lemomo-ai/lemo-opuscar.git
cd lemo-opuscar
claude
Films go into films/<name>/ inside the repo.
Then just say what you want. For example:
Make a 45-second film in the watercolor style about the coffee farm my family runs. Warm female narrator.
Or in Chinese:
用油画厚涂风格做一支 30 秒的片子，讲我家那只每天在窗台等我下班的橘猫。
You can name the style in English or Chinese. The style index lists them all. The agent reads three guides and works like a small studio.
You can watch the gallery and download the OPUSCAR 98 film at the project's GitHub Pages site. The repository is at github.com/lemomo-ai/lemo-opuscar.
Final Thoughts
Lemo-Opuscar is best for developers who are already using coding agents and want to explore what those agents can do beyond typical software tasks. It's also for anyone curious about code-generated video as an alternative to prompt-based generation. The tradeoff is clear: you get reproducibility and full control over the source, but you need to be comfortable working with an agent and the specific styles it supports. If you're looking for a push-button video generator, this isn't it. If you want to see how far you can take code as a medium for film, this is a solid place to start.
Follow @githubprojects for more developer tools and open source projects.