streaming-tts-elevenlabs

name	streaming-tts-elevenlabs
description	Streaming text-to-speech via ElevenLabs WebSocket API with real-time audio generation
category	voice

ElevenLabs Streaming TTS

Low-latency streaming text-to-speech using the ElevenLabs WebSocket Streaming API. Maintains a persistent WebSocket connection per session and streams audio chunks as MP3 data as soon as each sentence boundary is reached, enabling near-instant audio playback.

Setup

Set ELEVENLABS_API_KEY in your environment or agent secrets store.

Features

• Persistent WebSocket connection per session (no per-request HTTP overhead)
• Real-time audio streaming: ElevenLabs generates and sends audio as tokens arrive
• Automatic sentence-boundary flushing (., ?, !) for low-latency output
• Configurable voice via ElevenLabs voice ID
• Configurable model (eleven_turbo_v2 for low latency, eleven_multilingual_v2 for quality)
• Voice settings: stability, similarity boost, style, speaker boost
• MP3 output at 44.1 kHz / 128 kbps

Configuration

In agent.config.json:

{
  "voice": {
    "tts": "elevenlabs"
  }
}

Provider-specific options via providerOptions:

{
  "voice": {
    "tts": "elevenlabs",
    "providerOptions": {
      "voiceId": "21m00Tcm4TlvDq8ikWAM",
      "modelId": "eleven_turbo_v2",
      "stability": 0.5,
      "similarityBoost": 0.75,
      "style": 0.0,
      "useSpeakerBoost": true
    }
  }
}

Events

Event	Payload	Description
audio_chunk	EncodedAudioChunk	MP3 audio buffer ready for playback
utterance_complete	{ text, durationMs }	ElevenLabs signalled final audio generation done
cancelled	{ remaining: string }	Session was cancelled; remaining text not rendered
error	Error	WebSocket or synthesis error
close	—	Session fully terminated