A Python script to generate summaries (Claude), podcasts (Whisper), and videos (RunwayML or Luma AI) from annoyingly long YouTube content.

• Original video: https://www.youtube.com/watch?v=_K-L9uhsBLM
• Summary: https://dl.dropbox.com/scl/fi/mdkbglfbs4m9ydeo9a2k7/video-_K-L9uhsBLM.mp4?rlkey=3wrowryg9gio1walaxhdbp2is&dl=0

• Generate concise summaries of YouTube videos
• Create engaging podcast scripts with multiple voices
• Generate AI-powered videos with synchronized podcast audio
• Support for multiple languages
• Multiple transcription options
• Multiple video generation providers

1. Clone the repository:

git clone https://github.com/sliday/ytsum.git
cd ytsum

2. Install dependencies:

pip install -r requirements.txt

3. Install FFmpeg (required for audio/video processing):
   • macOS: brew install ffmpeg
   • Ubuntu/Debian: sudo apt-get install ffmpeg
   • Windows: Download from FFmpeg website

Create a .env file with your API keys:

ANTHROPIC_API_KEY=your_claude_api_key
OPENAI_API_KEY=your_openai_api_key
LUMAAI_API_KEY=your_lumaai_api_key
RUNWAYML_API_SECRET=your_runwayml_api_key
REPLICATE_API_TOKEN=your_replicate_api_key

python ytsum.py "https://www.youtube.com/watch?v=VIDEO_ID"

python ytsum.py --podcast "https://www.youtube.com/watch?v=VIDEO_ID"

# Using Luma AI (faster, recommended)
python ytsum.py --podcast --lumaai "https://www.youtube.com/watch?v=VIDEO_ID"

# Using RunwayML
python ytsum.py --podcast --runwayml "https://www.youtube.com/watch?v=VIDEO_ID"

• --language: Specify output language (default: english)
• --ignore-subs: Force transcription even when subtitles exist
• --fast-whisper: Use Fast Whisper for transcription (faster)
• --whisper: Use OpenAI Whisper for transcription (more accurate)
• --replicate: Use Replicate's Incredibly Fast Whisper

All output files are saved in the out directory:

• summary-{video_id}.txt: Text summary
• podcast-{video_id}.txt: Podcast script
• podcast-{video_id}.mp3: Podcast audio
• video-{video_id}.mp4: Final video with podcast audio

The tool supports two AI video generation providers:

• Faster generation times
• High-quality cinematic videos
• Supports camera movements and scene transitions
• Maintains visual consistency
• Optional image input for style reference

• High-quality video generation
• Requires input image
• Longer processing times
• Professional-grade output

Both providers:

1. Generate base images using Flux AI
2. Create video segments based on podcast content
3. Combine segments with audio
4. Support custom duration and aspect ratio

1. Fast Whisper (Default)

   • Quick transcription
   • Good accuracy
   • No API key required

2. OpenAI Whisper

   • High accuracy
   • Slower processing
   • Requires OpenAI API key

3. Replicate Whisper

   • Fastest option
   • Good accuracy
   • Requires Replicate API key

Run the test suite:

python test_ytsum.py

Run specific test groups:

# Run Luma AI tests only
pytest -v -m luma

# Run RunwayML tests only
pytest -v -m runway

• anthropic: Claude API for text generation
• openai: Whisper API for transcription and TTS
• lumaai: Luma AI for video generation (recommended)
• runwayml: RunwayML for video generation
• replicate: Flux AI for image generation
• ffmpeg-python: Audio/video processing
• colorama: Terminal output formatting
• pytest: Testing framework

1. Fork the repository
2. Create a feature branch
3. Commit your changes
4. Push to the branch
5. Create a Pull Request

This project is licensed under the MIT License - see the LICENSE file for details.