LMNTTTS

Convert text to speech using LMNT's Blizzard model via a simple REST API.

Use LMNTTTS when you want low-latency, high-quality speech synthesis from LMNT’s Blizzard model via a simple REST call. Each synthesize() request returns the complete audio as a Blob — no WebSocket management required. Voices from LMNT’s catalog and instant voice cloning are both supported through the voice option.

Prerequisites

An LMNT API key or a CompositeVoice proxy server
No additional dependencies required. LMNTTTS uses native fetch internally.

Basic setup

import {
  CompositeVoice,
  NativeSTT,
  AnthropicLLM,
  LMNTTTS,
  BrowserAudioOutput,
} from 'composite-voice';

const voice = new CompositeVoice({
  providers: [
    new NativeSTT(),
    new AnthropicLLM({
      proxyUrl: '/api/proxy/anthropic',
      model: 'claude-haiku-4-5',
    }),
    new LMNTTTS({
      proxyUrl: '/api/proxy/lmnt',
      voice: 'leah',
      model: 'blizzard',
      format: 'mp3',
    }),
    new BrowserAudioOutput(),
  ],
});

await voice.initialize();
await voice.startListening();

Configuration options

Option	Type	Default	Description
`apiKey`	`string`	—	LMNT API key (direct mode)
`proxyUrl`	`string`	—	Proxy server URL (recommended for production)
`voice`	`string`	required	Voice ID from `GET /v1/ai/voice/list` or a cloned voice
`model`	`string`	`'blizzard'`	The speech model to use
`format`	`string`	`'mp3'`	Output format: `mp3`, `wav`, `aac`, `ulaw`, `webm`, `pcm_s16le`, `pcm_f32le`
`sampleRate`	`number`	API default	Output sample rate: `8000`, `16000`, or `24000` Hz
`language`	`string`	auto-detect	Two-letter ISO 639-1 code (e.g. `en`), or `auto`
`temperature`	`number`	API default	Expressiveness — lower is more neutral, higher is more dynamic
`topP`	`number`	API default	Stability — lower is more consistent, higher is more flexible
`endpoint`	`string`	—	Custom API endpoint URL
`maxRetries`	`number`	`3`	Retry count for failed requests

Available voices

List voices with LMNT’s GET /v1/ai/voice/list endpoint, or create your own with instant voice cloning. The catalog voice leah is a good starting point.

Output formats

Format	Use case
`mp3`	96kbps MP3; good compression, wide browser support (default)
`wav`	16-bit PCM in a WAV container; uncompressed, highest quality
`aac`	Optimized for mobile devices
`ulaw`	8-bit G711 µ-law with WAV header; telephony
`webm`	WebM container with Opus codec
`pcm_s16le` / `pcm_f32le`	Raw PCM without a container; for custom audio pipelines

Complete example

import {
  CompositeVoice,
  MicrophoneInput,
  DeepgramSTT,
  AnthropicLLM,
  LMNTTTS,
  BrowserAudioOutput,
} from 'composite-voice';

const tts = new LMNTTTS({
  proxyUrl: '/api/proxy/lmnt',
  voice: 'leah',
  model: 'blizzard',
  format: 'mp3',
  language: 'en',
  temperature: 0.7,
});

const voice = new CompositeVoice({
  providers: [
    new MicrophoneInput(),
    new DeepgramSTT({ proxyUrl: '/api/proxy/deepgram' }),
    new AnthropicLLM({
      proxyUrl: '/api/proxy/anthropic',
      model: 'claude-haiku-4-5',
    }),
    tts,
    new BrowserAudioOutput(),
  ],
});

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

await voice.initialize();
await voice.startListening();

Expressiveness controls

temperature — influences how expressive and emotionally varied the speech becomes. Lower values (like 0.3) create more neutral, consistent speaking styles; higher values (like 1.0) allow more dynamic emotional range.
topP — controls the stability of the generated speech. A lower value (like 0.3) produces more consistent, reliable speech; a higher value (like 0.9) gives more flexibility but might occasionally produce unusual intonations.
language — Blizzard supports 31 languages. Specifying the language is recommended for faster generation; omit it (or use auto) for automatic detection.

Tips

LMNTTTS is REST-based, not streaming. The full audio Blob is returned after the API processes the entire input (max 5000 characters per request). For real-time streaming, consider DeepgramTTS.
LMNT also offers a full-duplex streaming WebSocket API (speech sessions); this provider intentionally uses the simpler REST endpoint, and streaming support may be added later.
Use proxyUrl in production so your API key stays server-side — set lmntApiKey in your proxy config.