Voicevox

You can run a VOICEVOX-compatible speech synthesis server through MCP (Model Context Protocol) to power voice-enabled agents and scripts. This server lets you synthesize speech using VOICEVOX engines and COEIROINK via a unified MCP interface, enabling easy integration with your client apps and automation workflows.

How to use

Connect your MCP client to the VOICEVOX MCP Server to synthesize speech from text. Use the HTTP endpoint when you want to access the VOICEVOX SSE stream from a remote or local VOICEVOX engine. If you prefer running VOICEVOX locally via Docker, start the local stdio MCP channel that launches the VOICEVOX-compatible container and exposes the necessary environment variables. Ensure VOICEVOX engine is running and accessible at the configured API URL and speaker ID.

How to install

Prerequisites you need to have before installation:

Prerequisites for Windows environment include Node.js 18 or newer, VOICEVOX ENGINE (running locally, e.g., at http://localhost:50000 or similar), and VLC media player (path must be in your system PATH).

For Docker on WSL2, you need Docker and Docker Compose, WSL2, VOICEVOX ENGINE (local or in Docker), and a Linux environment with libraries installed: libsdl2-dev pulseaudio-utils pulseaudio. Also ensure you can access /mnt/wslg.

Install and configure the server with the following steps.

Clone the project and install dependencies.

How to install (continued)

git clone https://github.com/Dosugamea/voicevox-mcp-server.git
cd voicevox-mcp-server
npm install

VOICEVOX_API_URL=http://localhost:50021
VOICEVOX_SPEAKER_ID=1

How to install (Windows/Docker setup)

If you plan to use Docker, you don’t need to run the server from your editor. The Docker setup runs in stdio mode.

How to configure and start

Create a local environment file and adjust the VOICEVOX settings to match your environment. Copy the example to a real .env and set the VOICEVOX API URL and speaker ID as needed.

On Windows, build and start the server from your terminal (in the project root):

npm run build
npm start

Available tools

voicevox

MCP tool that interfaces with VOICEVOX ENGINE via its API URL and speaker ID to synthesize speech through MCP.