ReadMe

This notebook transcribes YouTube videos using WhisperX

Prerequisites

1. pytubefix - For downloading YouTube videos
2. WhisperX - For high-accuracy transcription (https://github.com/m-bain/whisperX)
   • If using CUDA 11.x: Install ctranslate2==3.24.0
   • Install numpy<2 if required
   • Requires GPU with CUDA support

Getting Started

1. Install WhisperX: Please refer to https://github.com/m-bain/whisperX

2. Install other dependencies:

   pip install -r requirements.txt

3. Create a .env file with your API keys:

   OPENAI_API_KEY=your_openai_key
   HF_AUTH_TOKEN=your_huggingface_token
   

Running the Script

# For YouTube videos
python transcribe_video.py --youtube https://www.youtube.com/watch?v=ZEyPHhBKgJ4

# For local audio files  
python transcribe_video.py --audio path/to/audio.mp3 

# transcribe-only
 python transcribe_video.py --audio path/to/audio.mp3 --transcribe-only

# For summarize text, will use general summary prompt combine summary + QA
python summarize.py "script.txt/.md" 

# For summarize earning call
python summarize.py "script.txt/.md" --transcript_type earning_transcript

Convert PDF

magic-pdf -p "path_to_pdf.pdf" -o "out_put_path" -m auto