For NVIDIA GPUs with CUDA support:

# Check if CUDA is available
python gpu.py

# If CUDA is not detected, install PyTorch with CUDA support:
pip uninstall torch torchvision torchaudio
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

Verify GPU setup:

import torch
print(f"CUDA available: {torch.cuda.is_available()}")
print(f"GPU name: {torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'N/A'}")

The app automatically detects your hardware and suggests compatible models:

• ✅ Green: Fully compatible
• ⚠️ Yellow: Compatible but may be slow
• ❌ Red: Not recommended for your hardware

• Real-time Progress: Live percentage and estimated time
• Console Output: Detailed processing logs
• System Monitoring: GPU/CPU usage indicators

• File Validation: Automatic format and size checking
• Size Limits: 500MB maximum file size
• Batch Processing: Process multiple files (planned feature)

🚀 Tiny - Best for:

• Quick rough transcripts
• Resource-constrained systems
• Real-time applications
• Testing and development

⚡ Base - Best for:

• General everyday transcription
• Good balance of speed and accuracy
• Most users' needs
• Default choice for beginners

📊 Small - Best for:

• Professional transcription needs
• When accuracy matters more than speed
• Business documentation
• Podcasts and interviews

🎯 Medium - Best for:

• High-accuracy requirements
• Professional content creation
• Legal or medical transcriptions
• Important business meetings

🏆 Large - Best for:

• Maximum accuracy requirements
• Critical professional content
• Academic research
• When every word counts perfectly

⚡ Turbo - Best for:

• Production environments
• Best overall choice
• Optimized large-v3 performance
• Recommended for most users

Problem: Whispr shows "CPU Only Mode" despite having an NVIDIA GPU

Solutions:

1. Check NVIDIA Drivers:

   nvidia-smi

   If this fails, update your NVIDIA drivers

2. Reinstall PyTorch with CUDA:

   pip uninstall torch torchvision torchaudio
   pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

3. Run GPU Diagnostic:

   python gpu.py

4. Restart after driver installation

Problem: Model download fails or is very slow

Solutions:

1. Check internet connection and firewall settings

2. Manual model download:

   import whisper
   model = whisper.load_model("base")  # Replace with desired model

3. Clear model cache:

   • Delete the whisper_models folder and restart

4. Use smaller model temporarily:

   • Switch to "tiny" or "base" for faster download

Problem: "Error getting audio duration" or processing failures

Solutions:

1. Install/Update FFmpeg:

   # Windows (using chocolatey)
   choco install ffmpeg
   
   # macOS (using homebrew)
   brew install ffmpeg
   
   # Ubuntu/Debian
   sudo apt update && sudo apt install ffmpeg

2. Check audio file integrity:

   • Try a different audio file
   • Convert to WAV format as a test

3. Update librosa:

   pip install --upgrade librosa

Problem: Cannot write to uploads or transcripts folders

Solutions:

1. Check folder permissions:

   # Create directories manually
   mkdir uploads transcripts whisper_models
   chmod 755 uploads transcripts whisper_models

2. Run with appropriate permissions:

   • On Windows: Run terminal as Administrator
   • On macOS/Linux: Check file ownership and permissions

Problem: Cannot access web interface or features not working

Solutions:

1. Check if server is running:

   # Should show server startup messages
   python app.py

2. Try different port:

   # In app.py, change the last line to:
   app.run(debug=True, host='0.0.0.0', port=5001)

3. Clear browser cache and cookies

4. Try different browser (Chrome, Firefox, Safari)

5. Check firewall settings - ensure port 5000 is not blocked

1. Use GPU acceleration (up to 10x faster than CPU)
2. Choose appropriate model (Tiny/Base for speed, Medium/Large for accuracy)
3. Close unnecessary applications to free up system resources
4. Use SSD storage for faster file I/O
5. Ensure adequate cooling to prevent thermal throttling