YouTube Video Transcriber

A Python script that automatically downloads YouTube videos and generates text transcriptions using OpenAI's Whisper model.

Features

• Downloads audio from YouTube videos using yt-dlp
• Transcribes audio using OpenAI's Whisper model
• Saves transcriptions as formatted text files
• Handles temporary file management
• Supports various audio formats through FFmpeg
• Processes transcriptions into readable sentences

Prerequisites

Before running this script, make sure you have the following installed:

pip install -r requirements.txt

Required packages:

• yt-dlp
• whisper
• librosa
• ffmpeg (system dependency)

Installation

1. Clone this repository:

git clone https://github.com/Khalid1G/Whisper-YouTube-Transcription.git
cd Whisper-YouTube-Transcription

2. Install the required packages:

pip install -r requirements.txt

3. Install FFmpeg (if not already installed):

• On Ubuntu/Debian: sudo apt-get install ffmpeg
• On macOS: brew install ffmpeg
• On Windows: Download from FFmpeg official website

Usage

1. Run the script:

python transcriber.py

2. Enter the YouTube video ID when prompted. The video ID is the part after v= in the YouTube URL. For example, for https://www.youtube.com/watch?v=dQw4w9WgXcQ, the ID is dQw4w9WgXcQ

3. The script will:

   • Download the audio from the YouTube video
   • Process the audio using Whisper
   • Save the transcription to Youtube/[video_id]_transcription.txt

Output

The transcription will be saved in the Youtube directory with the filename pattern: [video_id]_transcription.txt

The output text is formatted with one sentence per line for better readability.

Project Structure

.
├── transcriber.py          # Main script
├── yt_transcription.ipynb  # Notebook script
├── requirements.txt        # Python dependencies
├── Youtube/                # Directory for transcription outputs
└── README.md               # This file

Performance

• Transcription time varies based on:
  • Video length
  • Computer specifications
  • Chosen Whisper model ("base" in this implementation)

Limitations

• Requires stable internet connection for YouTube downloads
• Processing time depends on video length and hardware capabilities
• Accuracy depends on audio quality and Whisper model used

Contributing

Feel free to open issues or submit pull requests with improvements.

License

MIT License

Acknowledgments

• OpenAI Whisper
• yt-dlp