MCP Coqui TTS Server

A Model Context Protocol (MCP) server that provides text-to-speech synthesis capabilities using Coqui TTS, including voice cloning support.

Features

• Text-to-Speech Synthesis: Convert text to natural-sounding speech
• Multiple Models: Support for various TTS models and languages
• Voice Cloning: Clone voices from audio samples using XTTS models
• Long Text Support: Automatic chunking for longer texts
• Customizable Output: Control speed, speaker, and language settings

Prerequisites

Before using this MCP server, you need to install Coqui TTS:

pip install TTS

For voice cloning and concatenation features, you'll also need ffmpeg:

# macOS
brew install ffmpeg

# Ubuntu/Debian
sudo apt-get install ffmpeg

# Windows (using chocolatey)
choco install ffmpeg

Installation

From npm

npm install -g @s.lfr/mcp-coqui-tts

From Source

git clone https://github.com/yourusername/mcp-coqui-tts.git
cd mcp-coqui-tts
npm install
npm link

Usage

With Claude Desktop

Add to your Claude Desktop configuration (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "coqui-tts": {
      "command": "npx",
      "args": ["@s.lfr/mcp-coqui-tts"]
    }
  }
}

Or if installed from source:

{
  "mcpServers": {
    "coqui-tts": {
      "command": "node",
      "args": ["/path/to/mcp-coqui-tts/index.js"]
    }
  }
}

Available Tools

1. speak

Convert text to speech with customizable parameters.

Parameters:

• text (required): The text to convert to speech
• model: TTS model to use (default: "tts_models/en/ljspeech/tacotron2-DDC")
• output_path: Where to save the audio file
• speaker_idx: Speaker index for multi-speaker models
• language_idx: Language index for multi-language models
• speed: Speed factor (1.0 is normal speed)

Example:

{
  "text": "Hello, this is a test of the text to speech system.",
  "model": "tts_models/en/ljspeech/tacotron2-DDC",
  "output_path": "/tmp/output.wav",
  "speed": 1.2
}

2. list_models

List all available TTS models.

Parameters: None

3. synthesize_long_text

Synthesize longer texts with automatic chunking and concatenation.

Parameters:

• text (required): The long text to convert
• model: TTS model to use
• output_path: Where to save the final audio
• chunk_size: Maximum characters per chunk (default: 500)

Example:

{
  "text": "This is a very long text that will be automatically split into chunks...",
  "output_path": "/tmp/long_speech.wav",
  "chunk_size": 500
}

4. clone_voice

Clone a voice from an audio sample (requires XTTS model).

Parameters:

• text (required): Text to speak in the cloned voice
• reference_audio (required): Path to reference audio for voice cloning
• output_path: Where to save the output
• language: Language code (default: "en")

Example:

{
  "text": "This will be spoken in the cloned voice.",
  "reference_audio": "/path/to/sample.wav",
  "output_path": "/tmp/cloned_voice.wav",
  "language": "en"
}

Popular TTS Models

English Models

• tts_models/en/ljspeech/tacotron2-DDC - High quality English TTS
• tts_models/en/ljspeech/fast_pitch - Fast English TTS
• tts_models/en/vctk/vits - Multi-speaker English (110 speakers)

Multilingual Models

• tts_models/multilingual/multi-dataset/xtts_v2 - Supports voice cloning
• tts_models/multilingual/multi-dataset/your_tts - Multilingual with voice cloning

Other Languages

Run list_models to see all available models for different languages.

Deployment on Smithery

To deploy this MCP server on Smithery:

1. Fork this repository
2. Connect your GitHub account to Smithery
3. Create a new MCP server on Smithery
4. Select this repository
5. Deploy

The server will be automatically available for use with any MCP-compatible client.

Development

Running Locally

npm start

Testing

You can test the server using the MCP inspector:

npx @modelcontextprotocol/inspector node index.js

Troubleshooting

Common Issues

1. "tts: command not found"

   • Make sure Coqui TTS is installed: pip install TTS
   • Ensure Python/pip binaries are in your PATH

2. "ffmpeg: command not found"

   • Install ffmpeg for your operating system (see Prerequisites)

3. Model download fails

   • First run may take time as models are downloaded
   • Check internet connection
   • Ensure sufficient disk space (~1-5GB per model)

4. Voice cloning not working

   • Requires XTTS v2 model
   • Reference audio should be clear, 5-10 seconds long
   • WAV format recommended for reference audio

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

MIT

Acknowledgments

• Coqui TTS for the excellent TTS library
• Model Context Protocol for the MCP SDK

Support

For issues and questions, please open an issue on GitHub.