Coar231/tfgridnet-demo

🎤 TF-GridNet Speech Enhancement Demo

Real-time speech enhancement powered by TF-GridNet deep learning model.

Features

• —🎵 Upload audio files (WAV, MP3, FLAC, OGG)
• —🎤 Record audio directly from microphone
• —✨ Real-time speech enhancement
• —📊 Performance metrics (latency, RTF)
• —📉 Waveform visualization
• —⬇️ Download enhanced audio

Model Info

• —Architecture: TF-GridNet
• —Sample Rate: 8kHz
• —Parameters: 2.3M
• —Tasks: Speech Enhancement, Separation, Denoising, Dereverberation

Usage

1. 1.Upload an audio file or record from microphone
2. 2.Click "🚀 Enhance Audio" button
3. 3.Listen to the enhanced result
4. 4.Download the improved audio

Technical Details

The model uses TF-GridNet architecture with:

• —STFT window size: 128
• —LSTM hidden units: 128
• —Attention heads: 4
• —3 processing layers

Built with Streamlit and PyTorch.