Beta — This project is under active development. Expect rough edges.
Rust native GUI client for Qwen3-TTS voice synthesis.
Built with iced, communicates with a Python FastAPI inference server over HTTP.
- Voice cloning (single and multi-speaker)
- Voice design from text descriptions
- Custom voice with built-in speakers
- Built-in audio recording and playback
- Local Whisper transcription for reference audio
- Server lifecycle management (auto-start, auto-terminate)
| Feature | Status |
|---|---|
| Voice clone | Verified |
| Clone with upload | Verified |
| Multi-speaker clone | Verified |
| Voice design | Verified |
| Custom voice | Verified |
| GUI polish | In progress |
- Nix with flakes enabled
- NVIDIA GPU with CUDA support (tested with RTX 4070)
nix develop
cargo runThe GUI automatically spawns the Python TTS server on startup. On first run, model weights are downloaded from HuggingFace (~3.5 GB).
This project has only been tested on NixOS. It may or may not work on other Linux distributions, macOS, or Windows.
MIT