ElevenLabsTTS

Use ElevenLabsTTS for high-fidelity voice cloning and expressive synthesis. Text streams over a WebSocket connection using the ElevenLabs stream-input protocol, and audio chunks arrive incrementally for low-latency playback.

Prerequisites

• An ElevenLabs API key or a CompositeVoice proxy server
• A voice ID from the ElevenLabs Voice Library or your cloned voices
• No additional peer dependencies required

Basic setup

import { CompositeVoice, DeepgramSTT, AnthropicLLM, ElevenLabsTTS } from '@lukeocodes/composite-voice';

const voice = new CompositeVoice({
  stt: new DeepgramSTT({ proxyUrl: '/api/proxy/deepgram' }),
  llm: new AnthropicLLM({
    proxyUrl: '/api/proxy/anthropic',
    model: 'claude-haiku-4-5',
  }),
  tts: new ElevenLabsTTS({
    proxyUrl: '/api/proxy/elevenlabs',
    voiceId: '21m00Tcm4TlvDq8ikWAM',
    modelId: 'eleven_turbo_v2_5',
    outputFormat: 'pcm_24000',
  }),
});

await voice.start();

Configuration options

Models

Output formats

Complete example

import { CompositeVoice, DeepgramSTT, AnthropicLLM, ElevenLabsTTS } from '@lukeocodes/composite-voice';

const tts = new ElevenLabsTTS({
  proxyUrl: '/api/proxy/elevenlabs',
  voiceId: '21m00Tcm4TlvDq8ikWAM',
  modelId: 'eleven_multilingual_v2',
  stability: 0.7,
  similarityBoost: 0.9,
  outputFormat: 'pcm_24000',
});

const voice = new CompositeVoice({
  stt: new DeepgramSTT({ proxyUrl: '/api/proxy/deepgram' }),
  llm: new AnthropicLLM({
    proxyUrl: '/api/proxy/anthropic',
    model: 'claude-haiku-4-5',
  }),
  tts,
  logging: { enabled: true, level: 'debug' },
});

voice.on('tts:start', () => console.log('Speaking...'));
voice.on('tts:end', () => console.log('Done speaking'));

await voice.start();

Voice cloning controls

ElevenLabs provides two parameters that control how closely the synthesized voice matches the original sample:

• Stability (0 to 1) — Higher values produce more consistent, predictable output. Lower values add variation and expressiveness. Start at 0.5 and adjust to taste.
• Similarity boost (0 to 1) — Higher values match the original voice more closely. Lower values allow more creative variation. Start at 0.75.

const tts = new ElevenLabsTTS({
  proxyUrl: '/api/proxy/elevenlabs',
  voiceId: 'your-cloned-voice-id',
  stability: 0.3,          // More expressive
  similarityBoost: 0.9,    // Stay close to the original voice
});

Streaming protocol

ElevenLabsTTS uses the ElevenLabs stream-input protocol with BOS (Beginning of Stream) and EOS (End of Stream) messages. The SDK handles this automatically. For standalone use:

const tts = new ElevenLabsTTS({
  proxyUrl: '/api/proxy/elevenlabs',
  voiceId: '21m00Tcm4TlvDq8ikWAM',
});

await tts.initialize();
await tts.connect();       // Opens WebSocket, sends BOS with voice settings

tts.onAudio((chunk) => {
  // chunk.data contains PCM or MP3 audio
});

tts.sendText('Hello, ');
tts.sendText('world!');
await tts.finalize();      // Sends EOS, flushes remaining audio
await tts.disconnect();

Tips

• Find voice IDs in the ElevenLabs Voice Library or via the list voices API endpoint.
• Use eleven_turbo_v2_5 for the lowest latency. Use eleven_multilingual_v2 when you need non-English languages.
• For telephony applications, use ulaw_8000 output format.
• Lower stability values work well for conversational, expressive speech. Higher values suit narration and formal content.

Further reading

• API reference
• Providers reference
• Getting started