Skip to content

README.md

Latest commit

 

History

History
103 lines (78 loc) · 3.77 KB

README.md

File metadata and controls

103 lines (78 loc) · 3.77 KB

S2ST-Distill Web Demo

A Gradio-based web interface for testing distilled speech-to-speech translation models.

Quick Start

# Install dependencies
pip install -r demo/requirements.txt

# Run the demo
python demo/app.py

# Open http://localhost:7860 in your browser

Features

  • 🎙️ Record Audio: Use your microphone to record speech
  • 📁 Upload Files: Support for WAV/MP3 audio files
  • 🌐 Multiple Languages: EN↔ZH, ZH↔FR translation pairs
  • 🔊 Auto Playback: Hear translations immediately
  • 📊 Real-time Status: See translation progress and metrics

Usage

  1. Select a language pair from the dropdown
  2. Record audio or upload a file
  3. Click Translate
  4. Listen to the translated result!

Running with Trained Models

Place your ONNX models in the models/ directory:

models/
├── en_zh/
│   └── model.onnx
├── zh_en/
│   └── model.onnx
├── zh_fr/
│   └── model.onnx
└── fr_zh/
    └── model.onnx

Then run the demo:

python demo/app.py --model-dir models

Command Line Options

--host       Host to bind to (default: 127.0.0.1)
--port       Port to listen on (default: 7860)
--share      Create a public Gradio link
--model-dir  Directory containing trained models

Demo Mode

If no trained models are found, the demo runs in mock mode:

  • UI is fully functional
  • Audio processing works (record, upload, playback)
  • "Translation" returns slightly modified audio (for UI testing)

This allows you to test the interface before training models.

Screenshots

┌─────────────────────────────────────────────────────────────┐
│                   🎤 S2ST-Distill Demo                      │
├─────────────────────────────────────────────────────────────┤
│  Language Pair: [English → Chinese ▼]                       │
├─────────────────────────────────────────────────────────────┤
│  ┌─────────────────┐    ┌─────────────────────────────────┐ │
│  │ 🎙️ Record       │    │ 🔊 Translated Audio             │ │
│  │                 │    │                                 │ │
│  │  [● Record]     │    │  [▶ Play]                       │ │
│  │                 │    │                                 │ │
│  │ — OR —          │    │ Status:                         │ │
│  │                 │    │ ✅ Translated successfully!     │ │
│  │ 📁 Upload       │    │ 📥 Input: 3.2s from microphone  │ │
│  │ [Choose File]   │    │ 🌐 Direction: English → Chinese │ │
│  └─────────────────┘    └─────────────────────────────────┘ │
│                                                             │
│  [🔄 Translate]                                             │
└─────────────────────────────────────────────────────────────┘

Requirements

  • Python 3.10+
  • Modern web browser with microphone access
  • For full functionality: PyTorch + ONNX Runtime

License

MIT License - see LICENSE for details.