Whisper Transcription

A web application for YouTube video transcription using OpenAI Whisper and MLX-Whisper for Apple Silicon. This project provides a simple web interface to transcribe YouTube videos and download SRT subtitle files.

Features

• 🎵 YouTube video transcription to SRT format
• 🌐 Web-based interface for easy use
• 📱 Responsive design
• 🔄 Cross-platform compatibility
• ⚡ Automatic model selection (MLX-Whisper for Apple Silicon, OpenAI Whisper for others)
• 📥 Direct SRT file download

Requirements

• Python 3.8+
• FFmpeg (for audio processing)
• Apple Silicon Macs: MLX-Whisper for optimal performance
• Other platforms: OpenAI Whisper with CUDA/CPU support

Installation

1. Clone the repository:

   git clone https://github.com/MinjaeKim09/whisperYTtoSRT.git
   cd whisperYTtoSRT

2. Install Python dependencies:

   pip install -r requirements.txt

3. Install FFmpeg:
   • macOS: brew install ffmpeg
   • Ubuntu/Debian: sudo apt install ffmpeg
   • Windows: Download from FFmpeg website

Usage

1. Start the application:

   python app.py

2. Open your browser and navigate to:

   http://localhost:8080
   

3. Enter a YouTube URL and select your preferred model size.
4. Click "Start Transcription" and wait for processing to complete.
5. Download the generated SRT file.

How It Works

1. Audio Extraction: Downloads audio from the provided YouTube URL using yt-dlp
2. Transcription: Processes the entire audio file using Whisper (batch processing)
3. SRT Generation: Converts the transcription into standard SRT subtitle format
4. Download: Provides the SRT file for download

Model Sizes

• Tiny: Fastest processing, less accurate
• Base: Good balance of speed and accuracy
• Small: Better accuracy than base
• Medium: Recommended balance (default)
• Large: Best accuracy, slower processing

Project Structure

whisperYTtoSRT/
├── app.py              # FastAPI web application
├── transcriber.py      # Core transcription logic
├── templates/          # HTML templates
├── requirements.txt    # Python dependencies
└── README.md           # This file

Technical Details

• Uses MLX-Whisper on Apple Silicon for optimal performance
• Falls back to OpenAI Whisper on other platforms
• Subprocess isolation ensures proper memory cleanup after transcription
• Temporary files are automatically cleaned up

License

MIT License - see LICENSE file for details.