F5-TTS WebSocket API Implementation

This is a WebSocket API implementation for F5-TTS that allows other programs to access TTS services through a WebSocket interface.

⚠️ Important Notes

Platform Support

• Windows: Fully tested and supported
• Linux: Not fully tested, may require modifications to scripts and paths
• macOS: Not tested

Work in Progress

Flash Attention 2 integration is currently under development and not functional. This feature will be added in future updates to improve performance.

Windows Installation (Recommended)

# Run the Windows installation script
setup.bat

This script will:

• Set up the required environment
• Install dependencies
• Configure the project for API usage

Usage on Windows

Starting the API Server

# Launch the WebSocket API server
tts-api.bat

The server will start on ws://localhost:8000 by default.

Testing the API

# Run the test script
test_api.bat

Linux Support

If you want to run this on Linux, you'll need to:

1. Convert the .bat scripts to .sh scripts
2. Modify file paths to use forward slashes
3. Adjust environment setup commands
4. Test the WebSocket implementation in your environment

We welcome contributions to improve Linux support!

WebSocket Protocol

Messages Format

Client to Server:

{
    "type": "tts_request",
    "data": {
        "text": "Text to synthesize",
        "reference_audio": "base64_encoded_audio",
        "config": {
            // Optional configuration parameters
        }
    }
}

Server to Client:

{
    "type": "tts_response",
    "data": {
        "audio": "base64_encoded_audio",
        "status": "success",
        "message": "Optional status message"
    }
}

Configuration

Custom TOML Configuration

The system uses TOML files for configuration. Here's an example structure:

# Model selection
model = "F5TTS_Base"

# Audio input/output settings
ref_audio = "reference/reference.wav"
ref_text = "Reference text for the audio"
gen_text = "Text to be generated"

# Optional settings
remove_silence = false
output_dir = "tests"
output_file = "output.wav"

Project Structure

F5-TTS/
├── api/                 # WebSocket API implementation
├── tests/              # Test files and output directory
├── reference/          # Reference audio files
├── setup.bat           # Windows installation script
├── test_api.bat        # Windows API testing script
├── tts-api.bat         # Windows server startup script
└── custom.toml         # Configuration file

Directory Usage:

• api/: Contains the WebSocket server implementation
• tests/: Used for storing test outputs and test scripts
• reference/: Stores reference audio files for TTS generation

Error Handling

The API includes robust error handling for common scenarios:

• Invalid audio format
• Missing configuration
• Server connection issues
• Processing errors

Contributing

Feel free to contribute to this project by:

1. Forking the repository
2. Creating a feature branch
3. Submitting a pull request

We especially welcome contributions for:

• Linux support improvements
• macOS compatibility
• Cross-platform testing

License

This project is licensed under the MIT License.