A PyQt6-based cross-platform desktop interface for CosyVoice3 intelligent voice synthesis, featuring a unique pixel-art aesthetic.
- OS: macOS 11+, Windows 10+, or Ubuntu 20.04+
- Python: 3.10
- RAM: 8GB (16GB Recommended)
- GPU: Optional (Supports NVIDIA CUDA & Apple Silicon MPS)
The application follows a clean layered structure to ensure stability and performance:
- Presentation: PyQt6 UI with retro pixel-art styling.
- Controller: Event handling and UI logic.
- Service: Business logic encapsulation.
- Worker: Asynchronous task processing via
QThread. - Core: CosyVoice3, PyTorch, and audio processing engine.
- Reference Audio: Select a clear
.wavfile via the BROWSE button. - Model Selection: Choose your preferred CosyVoice model from the dropdown.
- Synthesis: Enter your text (supports multiple languages) and adjust the Pitch (-12 to +12).
- Generate: Click GENERATE and wait for the progress bar to complete.
Access the MODEL DOWNLOAD page to manage your synthesis engines:
- View real-time download speed and progress.
- Download individual models or use DOWNLOAD ALL.
- Refresh to update current model statuses.
| Model Name | Size | Description |
|---|---|---|
| CosyVoice3-0.5B-2512 | ~1.2 GB | Latest flagship model (Recommended) |
| CosyVoice2-0.5B | ~980 MB | Balanced performance |
| CosyVoice-300M | ~600 MB | Lightweight and fast |
| CosyVoice-TTSFRD | ~550 MB | Optimized for fast response |
Built with PyQt6 + CosyVoice3