Delayed Streams UI Setup

A portable web front end for Kyutai's open-source TTS/STT models (unofficial, unsupported, personal project)

Disclaimer: This project is an independent front end for Kyutai's open-source models. It is not affiliated with or endorsed by Kyutai. This is my first time doing this sort of coding and GitHub work—it's a personal side project, provided as-is, with no guarantees or support. I hope someone finds it useful!

Quick Start (Windows)

1. Download or clone this repository.

2. Only use the setup scripts in the project root.

3. Open a Command Prompt or PowerShell window in the project directory.

4. Run the setup script:

   setup_windows.bat
   

5. Follow any prompts.

   • The script will automatically install Python (if needed), all dependencies, clone both the backend and UI repos, and set up a virtual environment.
   • If you see a warning about WSL (Windows Subsystem for Linux), follow the instructions to install it for backend (Rust STT/TTS) support.

6. To start the UI: Run the start script:

   start-ui.bat
   

   Then open http://localhost:5000 in your browser.

Notes:

• The script installs the CUDA (GPU) version of PyTorch by default. You need a compatible NVIDIA GPU and drivers for GPU acceleration.
• If you do not have a compatible GPU, you may need to manually install the CPU version of PyTorch.
• WSL is required for the backend (Rust STT/TTS server). If not installed, the script will prompt you with instructions.
• Do not use any install or setup scripts inside the /ui/ folder. All setup should be done from the project root.

---

Quick Start (Linux)

1. Download or clone this repository.

2. Only use the setup scripts in the project root.

3. Open a terminal in the project directory.

4. Run the setup script:

   bash setup_linux.sh
   

5. Follow the prompts to complete the setup.

6. To start the UI:

   cd delayed-streams-ui
   source tts-venv/bin/activate
   python run.py
   

   Then open http://localhost:5000 in your browser.

Notes:

• The Linux script installs the CUDA (GPU) version of PyTorch by default for GPU acceleration. You need a compatible NVIDIA GPU and drivers.
• If you do not have a compatible GPU, you may need to manually install the CPU version of PyTorch.
• Do not use any install or setup scripts inside the /ui/ folder. All setup should be done from the project root.

---

Backend and UI

• The setup scripts will clone both the backend (Kyutai's delayed-streams-modeling) and the UI (delayed-streams-ui) repositories.
• All dependencies are installed automatically.

---

Troubleshooting

• If you encounter issues with CUDA or GPU support, check your NVIDIA driver and CUDA toolkit installation.
• If WSL is not installed on Windows, follow the script's prompt or see https://aka.ms/wslinstall.
• For CPU-only PyTorch, see PyTorch Get Started.

Credits

• Kyutai for the TTS/STT models and backend: kyutai-labs/delayed-streams-modeling
• This front end is an independent project to provide a user-friendly interface for Kyutai's models.

License

MIT License (see LICENSE file)