Skip to main content
Glama

MCP FishAudio Server

by da-okazaki
websocket-streaming-example.md2.57 kB
# WebSocket Streaming Examples ## Overview Fish Audio MCP Server now supports WebSocket streaming for real-time text-to-speech generation. This enables low-latency audio streaming with the ability to play audio as it's being generated. ## WebSocket vs HTTP Streaming - **HTTP Streaming**: Downloads the entire audio file in chunks - **WebSocket Streaming**: Bidirectional real-time communication, audio plays as it's generated ## Usage Examples ### Basic WebSocket Streaming ``` User: "Generate speech with WebSocket: 'Hello from the future'" Claude: I'll use WebSocket streaming for low-latency speech generation. [Uses fish_audio_tts tool with websocket_streaming: true] Result: Audio generated via WebSocket and saved to file ``` ### Real-time Playback ``` User: "Stream and play in real-time: 'This is being spoken as it's generated'" Claude: I'll stream and play the audio in real-time using WebSocket. [Uses fish_audio_tts tool with websocket_streaming: true, realtime_play: true] Result: Audio streams and plays immediately as chunks arrive ``` ### WebSocket with Custom Voice ``` User: "Use WebSocket streaming with voice xyz123: 'Custom voice in real-time'" Claude: I'll use WebSocket streaming with your specified voice model. [Uses fish_audio_tts tool with websocket_streaming: true, reference_id: "xyz123"] Result: Real-time speech with custom voice via WebSocket ``` ## Configuration ### Enable WebSocket Streaming by Default ```json { "mcpServers": { "fish-audio": { "command": "npx", "args": ["@alanse/fish-audio-mcp-server"], "env": { "FISH_API_KEY": "your_api_key", "FISH_WEBSOCKET_STREAMING": "true", "FISH_REALTIME_PLAY": "true" } } } } ``` ## Technical Details ### WebSocket Endpoint - URL: `wss://api.fish.audio/v1/tts/live` - Protocol: WebSocket with MessagePack encoding ### Message Flow 1. Connect to WebSocket with Bearer token 2. Send `start` event with TTS parameters 3. Send `text` events with text chunks 4. Receive audio data in real-time 5. Send `stop` event to end session ### Supported Formats - `opus` (recommended for streaming) - `mp3` - `wav` - `pcm` ### Real-time Players - **macOS**: `ffplay` (opus/mp3) or `afplay` - **Windows**: `ffplay` - **Linux**: `mpv` or `ffplay` ## Benefits 1. **Low Latency**: Audio starts playing before full generation completes 2. **Memory Efficient**: No need to buffer entire audio file 3. **Interactive**: Can be used for real-time conversational AI 4. **Smooth Playback**: Continuous audio stream without gaps

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/da-okazaki/mcp-fish-audio-server'

If you have feedback or need assistance with the MCP directory API, please join our Discord server