Supertonic3 Mcp

@nextic-tech

Visit Server

4 hours ago

Local, on-device text-to-speech for Claude & Cursor. No API key, no cloud. Powered by Supertonic 3 — 10 voices, 31 languages, inline expression tags (<laugh>, <pause>, etc.). ~820ms synthesis on Apple Silicon. MIT licensed.

Overview Tools Comments

Overview

What it does

Give your AI assistant a voice — locally, with no API key and no cloud calls. supertonic3-mcp wraps Supertonic 3 on-device TTS as three MCP tools your AI assistant (Claude Desktop, Cursor, or any MCP client) can call directly.

Tools

Tool	Description
`speak`	Synthesize text to WAV. Returns file path + metadata.
`list_voices`	List built-in voices (M1–M5, F1–F5)
`list_expressions`	List inline expression tags with descriptions

Key features

No API key — runs entirely on your machine
No cloud — model weights cached locally (~400MB, downloads once)
10 voices — M1–M5, F1–F5
31 language codes — en, ko, ja, fr, and more
Inline expression tags — <laugh>, <pause>, <breath>, and 7 more
~820ms warm synthesis on Apple M3 Air
FastMCP + asyncio — inference serialized behind asyncio.Lock; playback runs outside the lock

Quick start

git clone https://github.com/nextic-tech/supertonic3-mcp
cd supertonic3-mcp
pip install -e .
supertonic3-mcp preload  # optional: pre-download model

Add to your MCP config:

{
  "mcpServers": {
    "supertonic3": {
      "command": "/absolute/path/to/supertonic3-mcp/.venv/bin/supertonic3-mcp",
      "args": []
    }
  }
}

License

Try in Playground

Server Config

{
  "mcpServers": {
    "supertonic3": {
      "command": "uvx",
      "args": [
        "supertonic3-mcp"
      ]
    }
  }
}

Build with ShipAny.