| 1 | # Parakeet ASR Demo |
| 2 | |
| 3 | Speech recognition using NVIDIA's Parakeet TDT model. |
| 4 | |
| 5 | ## Features |
| 6 | |
| 7 | - 🎙️ Speech-to-text transcription using NVIDIA Parakeet TDT model |
| 8 | - 📊 Real-time transcription with progress tracking |
| 9 | - 📝 Support for multiple audio formats (WAV, FLAC) |
| 10 | - 📈 Transcription history with export options |
| 11 | - 🎯 Optimized for both short and long audio files |
| 12 | - 💻 GPU acceleration support with fallback to CPU |
| 13 | |
| 14 | ## Setup |
| 15 | |
| 16 | 1. Install dependencies: |
| 17 | ```bash |
| 18 | pip install -r requirements.txt |
| 19 | ``` |
| 20 | |
| 21 | 2. Run the application: |
| 22 | ```bash |
| 23 | streamlit run app.py |
| 24 | ``` |
| 25 | |
| 26 | ## Requirements |
| 27 | |
| 28 | - Python 3.8+ |
| 29 | - NVIDIA GPU with CUDA support (strongly recommended for optimal performance) |
| 30 | - FFmpeg (for audio processing) |
| 31 | |
| 32 | ## Usage |
| 33 | |
| 34 | 1. Upload an audio file or record directly in the browser |
| 35 | 2. Wait for the model to process and transcribe |
| 36 | 3. View and export transcription results |
| 37 | |
| 38 | ## Notes |
| 39 | |
| 40 | - NVIDIA GPU with CUDA support is strongly recommended for optimal performance |
| 41 | - Long audio files (>8 minutes) will automatically use optimized settings |
| 42 | - Maximum recommended audio duration is 30 minutes |