Fish Audio

You deploy and run the Fish Audio MCP Server to enable natural language driven speech synthesis with Fish Audio’s TTS API inside your MCP-enabled workflows. This server provides two tools to your LLMs: one for generating speech from text and another for listing configured voice references, with flexible options for voice selection, streaming, and output formats.

How to use

Use the Fish Audio MCP Server by running it in your MCP environment and invoking its tools from your MCP-compatible client. You can generate speech from text, choose a voice by ID, name, or tag, and control streaming, output format, and latency to fit your application. You can also query the configured voice references to understand available voices and defaults.

How to install

Prerequisites: you need Node.js and npm installed on your machine. Ensure you have a Fish Audio API key to access the TTS service.

Option 1: Run the MCP server directly with npx without local installation.

npx @alanse/fish-audio-mcp-server

Option 2: Install the MCP server globally for easier reuse.

npm install -g @alanse/fish-audio-mcp-server

Configure environment variables and start the server via MCP settings as shown in the configuration examples.

Configuration and usage notes

Configuration is done through environment variables passed to the MCP server when it is invoked. The following environment variables control core behavior: FISH_API_KEY, FISH_MODEL_ID, FISH_REFERENCE_ID, FISH_REFERENCES, FISH_DEFAULT_REFERENCE, FISH_OUTPUT_FORMAT, FISH_STREAMING, FISH_LATENCY, FISH_MP3_BITRATE, FISH_AUTO_PLAY, and AUDIO_OUTPUT_DIR.

Single Voice Mode (Simple) uses a single default voice reference. Example settings show using the following env values when starting the server via MCP: FISH_API_KEY, FISH_MODEL_ID, FISH_REFERENCE_ID, FISH_OUTPUT_FORMAT, FISH_STREAMING, FISH_LATENCY, FISH_MP3_BITRATE, FISH_AUTO_PLAY, AUDIO_OUTPUT_DIR.

Multiple Voice Mode (Advanced) allows configuring multiple voice references and selecting a default reference. The environment variables include FISH_REFERENCES as a JSON array of references and FISH_DEFAULT_REFERENCE to designate the default voice.

If you configure multiple references, you can also provide the references individually with FISH_REFERENCE_1_ID, FISH_REFERENCE_1_NAME, and FISH_REFERENCE_1_TAGS, and so on for additional references.

Configuration snippets (examples)

{
  "mcpServers": {
    "fish_audio": {
      "command": "npx",
      "args": ["-y", "@alanse/fish-audio-mcp-server"],
      "env": {
        "FISH_API_KEY": "your_fish_audio_api_key_here",
        "FISH_MODEL_ID": "speech-1.6",
        "FISH_REFERENCE_ID": "your_voice_reference_id_here",
        "FISH_OUTPUT_FORMAT": "mp3",
        "FISH_STREAMING": "false",
        "FISH_LATENCY": "balanced",
        "FISH_MP3_BITRATE": "128",
        "FISH_AUTO_PLAY": "false",
        "AUDIO_OUTPUT_DIR": "~/.fish-audio-mcp/audio_output"
      }
    }
  }
}

Environment variables

The following environment variables are used to configure the MCP server. You should provide values appropriate for your setup.

FISH_API_KEY=your_fish_audio_api_key_here
FISH_MODEL_ID=speech-1.6
FISH_REFERENCE_ID=your_voice_reference_id_here
FISH_REFERENCES=[{'reference_id':'id1','name':'Alice','tags':['female','english']},{'reference_id':'id2','name':'Bob','tags':['male','japanese']}]
FISH_DEFAULT_REFERENCE=id1
FISH_OUTPUT_FORMAT=mp3
FISH_STREAMING=false
FISH_LATENCY=balanced
FISH_MP3_BITRATE=128
FISH_AUTO_PLAY=false
AUDIO_OUTPUT_DIR=~/.fish-audio-mcp/audio_output

Usage details

Tool 1: fish_audio_tts generates speech from text using Fish Audio's TTS API. It supports selecting a voice by reference_id, reference_name, or reference_tag, streaming options, output format, and latency. The tool prioritizes reference_id, then reference_name, then reference_tag, and finally the default when selecting a voice.

Tool 2: fish_audio_list_references lists all configured voice references along with their IDs, names, and tags, as well as the default reference.

Available tools

fish_audio_tts

Generates speech from text using Fish Audio's TTS API with voice selection, streaming, and formatting options.

fish_audio_list_references

Lists all configured voice references with IDs, names, and tags, plus the default reference.