bratishka

CLI tool to download YouTube videos, transcribe with Whisper, and generate AI-powered reports.

Features

• Download videos from YouTube using yt-dlp
• Native Rust transcription with whisper-rs (CUDA accelerated)
• Generate structured reports with AI (Grok, OpenAI, or Gemini)
• Smart caching - skip already-completed steps
• Multi-language report generation
• Auto-downloads Whisper model on first run

Requirements

• yt-dlp - Video downloader
• ffmpeg - Audio extraction
• NVIDIA GPU with CUDA (optional, falls back to CPU)
• One of: XAI_API_KEY, OPENAI_API_KEY, or GEMINI_API_KEY

Install dependencies

# macOS
brew install yt-dlp ffmpeg

# Arch Linux
sudo pacman -S yt-dlp ffmpeg

# Ubuntu/Debian
sudo apt install yt-dlp ffmpeg

Installation

From source

cargo install --git https://github.com/kuchmenko/bratishka-rs

From GitHub releases

Download the latest release for your platform from Releases.

Usage

# Basic usage (uses Grok by default)
export XAI_API_KEY=your-key
bratishka "https://youtube.com/watch?v=..."

# Use OpenAI
export OPENAI_API_KEY=your-key
bratishka "https://youtube.com/watch?v=..." -p openai

# Use Gemini
export GEMINI_API_KEY=your-key
bratishka "https://youtube.com/watch?v=..." -p gemini

# Force specific report language
bratishka "https://youtube.com/watch?v=..." -l en

# Force re-processing (ignore cache)
bratishka "https://youtube.com/watch?v=..." --force

Options

Arguments:
  <URL>  Video URL

Options:
  -l, --lang <LANG>          Report language (defaults to video's detected language)
  -p, --provider <PROVIDER>  AI provider [default: grok] [possible values: grok, openai, gemini]
  -f, --force                Force re-processing even if cached files exist
  -h, --help                 Print help

Output

Reports are cached in ~/.cache/bratishka/<url-hash>/ and include:

• video.* - Downloaded video
• audio.wav - Extracted audio
• transcript.json - Whisper transcription
• report_<provider>_<lang>.json - AI-generated report

Report structure

{
  "title": "Video title",
  "summary": "2-3 sentence summary",
  "duration_minutes": 45.5,
  "language": "en",
  "difficulty": "Intermediate",
  "topics": ["topic1", "topic2"],
  "key_takeaways": ["takeaway1", "takeaway2"],
  "chapters": [
    {
      "start_seconds": 0,
      "end_seconds": 180,
      "title": "Introduction",
      "summary": "Chapter summary"
    }
  ],
  "prerequisites": ["prerequisite1"],
  "target_audience": "Who this video is for"
}

License

MIT