home / mcp / blabber-mcp mcp server
Provides text-to-speech synthesis via OpenAI, offering voice and model options, audio formats, and optional playback.
Configuration
View docs{
"mcpServers": {
"pinkpixel-dev-blabber-mcp": {
"command": "node",
"args": [
"/full/path/to/blabber-mcp/build/index.js"
],
"env": {
"OPENAI_API_KEY": "YOUR_OPENAI_API_KEY",
"DEFAULT_TTS_VOICE": "nova",
"AUDIO_PLAYER_COMMAND": "xdg-open"
}
}
}
}Blabber-MCP is a local MCP server that gives your language models a voice by converting text to speech using OpenAI's Text-to-Speech API. You can choose from multiple voices and models, save the audio, and optionally playback the result automatically. This guide shows you how to configure, install, and use the server with an MCP client.
You will interact with Blabber-MCP through your MCP client by calling the text_to_speech tool. This tool sends input text and options to the Blabber-MCP server, which returns a generated audio file. You can specify the voice, model, and output format, and you can enable automatic playback after the file is saved.
Prerequisites: ensure you have Node.js and npm installed on your system.
1. Clone the Blabber-MCP project to a local directory.
2. Install dependencies.
3. Build the project to generate the runtime index.js.
4. Create the MCP client configuration block to run Blabber-MCP as a stdio server (details below).
5. Start or enable the MCP server integration in your environment using the configured command.
Environment variables and defaults control how Blabber-MCP operates. You can set an OpenAI API key, choose a default voice, and specify a system command to play audio. The server saves generated audio to a local path and can automatically playback the file if you enable it.
The built-in tool is text_to_speech. When you call it, you can provide the input text, optionally select a voice from alloy, echo, fable, onyx, nova, or shimmer, choose a model (tts-1 or tts-1-hd), and set the output format (mp3, opus, aac, flac). You can also request playback immediately after the file is created.
{
"mcpServers": {
"blabber_mcp": {
"command": "node",
"args": ["/full/path/to/blabber-mcp/build/index.js"],
"env": {
"OPENAI_API_KEY": "YOUR_OPENAI_API_KEY",
"AUDIO_PLAYER_COMMAND": "xdg-open",
"DEFAULT_TTS_VOICE": "nova"
},
"disabled": false,
"alwaysAllow": []
}
}
}You invoke the text_to_speech tool on the Blabber-MCP server from your MCP client. Include the input text and optional parameters as needed. The server processes the request, saves the audio to the output directory, and returns information about the saved file and the voice used.
The generated audio is saved to an output directory in the Blabber-MCP project folder. If playback is requested, the server uses the configured audio player command (for example, cvlc, vlc, mpv, ffplay, afplay, or xdg-open) to play the file after saving.
Voices you can choose from are alloy, echo, fable, onyx, nova, and shimmer. The available models are tts-1 (standard) and tts-1-hd (high-definition). Audio formats include mp3, opus, aac, and flac. If you do not specify a voice or format, sensible defaults are used.
Protect your OpenAI API key. Store it in the OPENAI_API_KEY environment variable and avoid exposing it in logs or client configurations.
Use absolute paths for the built runtime index when configuring the MCP server. The provided example uses the full path to the built index.js file in the blabber-mcp project.
Tool to synthesize speech from text using a specified voice, model, and output format. Returns saved audio file information and can trigger playback if requested.