README.md

metadata

title: Faster Whisper API
emoji: 🎤
colorFrom: blue
colorTo: purple
sdk: docker
sdk_version: latest
app_file: app.py
pinned: false

🎤 Faster Whisper API - Fixed Version

🆕 Latest Fixes Applied:

✅ Critical Bug Fixes:

• Fixed "name 'traceback' is not defined" error - Removed problematic traceback import
• Improved error handling - Better error messages and logging
• Enhanced CORS middleware - Better browser compatibility
• Added detailed logging - For easier debugging on Hugging Face Spaces

🔧 Performance Improvements:

• Better file validation - 25MB file size limit
• Enhanced VAD support - Voice Activity Detection with fallback
• Improved model loading - Better error handling during startup
• Added health check endpoint - For monitoring service status

🚀 Quick Start:

Health Check:

curl https://alaaharoun-faster-whisper-api.hf.space/health

Transcribe Audio (without VAD):

curl -X POST \
  -F "file=@audio.wav" \
  -F "language=en" \
  -F "task=transcribe" \
  https://alaaharoun-faster-whisper-api.hf.space/transcribe

Transcribe Audio (with VAD):

curl -X POST \
  -F "file=@audio.wav" \
  -F "language=en" \
  -F "task=transcribe" \
  -F "vad_filter=true" \
  -F "vad_parameters=threshold=0.5" \
  https://alaaharoun-faster-whisper-api.hf.space/transcribe

📊 Supported Parameters:

• file: Audio file (WAV, MP3, M4A, FLAC, OGG, WEBM)
• language: Language code (optional, e.g., "en", "ar", "es")
• task: "transcribe" or "translate" (default: "transcribe")
• vad_filter: Enable Voice Activity Detection (default: false)
• vad_parameters: VAD parameters (default: "threshold=0.5")

🔧 Response Format:

Success Response:

{
  "success": true,
  "text": "Transcribed text here",
  "language": "en",
  "language_probability": 0.95,
  "vad_enabled": false,
  "vad_threshold": null
}

Error Response:

{
  "error": "Error message",
  "error_type": "ExceptionType",
  "success": false
}

🛠️ Local Development:

# Install dependencies
pip install -r requirements.txt

# Run the server
python app.py

Or with uvicorn:

uvicorn app:app --host 0.0.0.0 --port 7860

📝 Important Notes:

• Maximum file size: 25MB
• Supported formats: WAV, MP3, M4A, FLAC, OGG, WEBM
• VAD support: Configurable threshold with fallback mechanism
• Language detection: Automatic if not specified
• Error handling: Detailed error messages for debugging

🔍 Troubleshooting:

Common Issues:

1. 500 Internal Server Error:

   • Check if the model is loaded properly
   • Verify file format and size
   • Check server logs for detailed error messages

2. VAD Issues:

   • The service will automatically fallback to standard transcription
   • Check VAD parameters format

3. File Upload Issues:

   • Ensure file size is under 25MB
   • Check file format compatibility

🌐 Service URLs:

• Main Service: https://alaaharoun-faster-whisper-api.hf.space
• Health Check: https://alaaharoun-faster-whisper-api.hf.space/health
• API Documentation: https://alaaharoun-faster-whisper-api.hf.space/docs

📈 Performance:

• Model: Whisper base model with int8 quantization
• Processing: Optimized for real-time transcription
• Memory: Efficient memory usage for Hugging Face Spaces
• Concurrency: Supports multiple concurrent requests

🔒 Security:

• CORS: Configured for cross-origin requests
• File Validation: Strict file type and size validation
• Error Handling: No sensitive information in error messages
• Authentication: Optional API token support (currently disabled)

📞 Support:

For issues or questions:

1. Check the health endpoint first
2. Review server logs for detailed error messages
3. Test with a simple audio file
4. Verify file format and size requirements