repo2brainrot

Turn a code repository into a TikTok-style doom-scroll video feed.

Point it at a repo, pick the most interesting parts, and turn the codebase into short-form episodes with narration, subtitles, code overlays, and a swipeable feed.

It works with GitHub Copilot CLI, GitHub Copilot in VS Code / agent mode, Claude Code, and other tools that support the open Agent Skills format.

What it produces

output/<repo-name>/
├── player.html
├── videos/ep01.mp4
├── videos/full_feed.mp4
├── scripts/
├── audio/
├── subs/
└── analysis.json

Each episode includes:

• AI narration with Kokoro TTS
• Word-level subtitles
• Syntax-highlighted code overlays from the real repo
• Vertical background footage
• Auto-splitting when a topic runs past 60 seconds

Use it with Copilot CLI

Install the whole repository as a plugin:

copilot plugin install elemarin/repo2brainrot

Then use the skill in a Copilot CLI session:

/repo2brainrot https://github.com/elemarin/repo2brainrot

Use it as a standalone skill

If you want to use the skill outside the plugin flow, copy skills/repo2brainrot/ into one of these supported skill locations:

• Copilot personal skills: ~/.copilot/skills/repo2brainrot/
• Claude personal skills: ~/.claude/skills/repo2brainrot/
• Generic Agent Skills location: ~/.agents/skills/repo2brainrot/

Or add it as a project skill inside a target repository:

• Copilot project skill: .github/skills/repo2brainrot/
• Claude project skill: .claude/skills/repo2brainrot/
• Generic project skill: .agents/skills/repo2brainrot/

The important part is the folder shape: one skill directory containing SKILL.md and its bundled resources.

Prerequisites

System dependencies

# ffmpeg (required for video assembly)
winget install Gyan.FFmpeg

# espeak-ng (required for Kokoro phonemizer)
# Windows: install from https://github.com/espeak-ng/espeak-ng/releases

Python dependencies

From a clone of this repository:

pip install -r skills/repo2brainrot/scripts/requirements.txt

Requires Python 3.11+ and enough disk space for PyTorch and the Kokoro model download.

Quick start

1. Add gameplay footage

Put 1-3 background clips in assets/.

2. Add a repo to analyze

git clone https://github.com/some/repo repos/my-repo

3. Run the skill

/repo2brainrot repos/my-repo

Copilot will analyze the repo, ask which categories you want, write scripts, generate videos, and build player.html.

Run the video pipeline manually

If you already have episode JSON files and want to render them directly from this repository:

python skills/repo2brainrot/scripts/generate_video.py \
  --scripts output/my-repo/scripts/ \
  --assets assets/ \
  --output output/my-repo/videos/ \
  --repo repos/my-repo

To serve the generated feed:

cd output/my-repo
python -m http.server 8888

Then open http://localhost:8888/player.html.

Repo layout

repo2brainrot/
├── plugin.json
├── README.md
├── .gitignore
├── skills/
│   └── repo2brainrot/
│       ├── SKILL.md
│       ├── player-template.html
│       ├── references/
│       │   ├── categories.md
│       │   ├── video-format.md
│       │   └── analysis-procedure.md
│       └── scripts/
│           ├── generate_video.py
│           ├── compress_videos.py
│           └── requirements.txt
├── assets/
├── repos/
└── output/

Tech stack

• Skill orchestration: Agent Skills + GitHub Copilot CLI
• TTS: Kokoro 82M
• Subtitle alignment: wav2vec2 via torchaudio
• Code rendering: Pillow
• Video assembly: FFmpeg

Demo

License

MIT