SpeechifyTTS

Convert text to speech using Speechify's Simba models via a simple REST API.

Use SpeechifyTTS when you want high-quality speech synthesis from Speechify’s Simba models via a simple REST call. Each synthesize() request returns the complete audio as a Blob — no WebSocket management required. Voices from Speechify’s catalog and instant voice cloning are both supported through the voiceId option.

Prerequisites

A Speechify API key or a CompositeVoice proxy server
No additional dependencies required. SpeechifyTTS uses native fetch internally.

Basic setup

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

const voice = new CompositeVoice({
  providers: [
    new NativeSTT(),
    new AnthropicLLM({
      proxyUrl: '/api/proxy/anthropic',
      model: 'claude-haiku-4-5',
    }),
    new SpeechifyTTS({
      proxyUrl: '/api/proxy/speechify',
      voiceId: 'geffen_32',
      model: 'simba-3.2',
      audioFormat: 'mp3',
    }),
    new BrowserAudioOutput(),
  ],
});

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

Configuration options

Option	Type	Default	Description
`apiKey`	`string`	—	Speechify API key (direct mode)
`proxyUrl`	`string`	—	Proxy server URL (recommended for production)
`voiceId`	`string`	required	Voice identifier from `GET /v1/voices` or a cloned voice
`model`	`string`	`'simba-english'`	`simba-english`, `simba-multilingual`, `simba-3.0`, `simba-3.2`
`audioFormat`	`string`	`'mp3'`	Output format: `mp3`, `wav`, `ogg`, `aac`
`language`	`string`	auto-detect	ISO 639-1 code with optional region (e.g. `en-US`)
`loudnessNormalization`	`boolean`	`false`	Normalize output loudness
`textNormalization`	`boolean`	`false`	Normalize numbers, dates, and abbreviations before synthesis
`endpoint`	`string`	—	Custom API endpoint URL
`maxRetries`	`number`	`3`	Retry count for failed requests

Available voices

List voices with Speechify’s GET /v1/voices endpoint, or create your own with instant voice cloning. The quickstart voice geffen_32 is a good starting point for the simba-3.2 model.

Output formats

Format	Use case
`mp3`	Good compression, wide browser support (default)
`wav`	Uncompressed, highest quality
`ogg`	Good compression, open format
`aac`	Optimized for mobile devices

Complete example

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

const tts = new SpeechifyTTS({
  proxyUrl: '/api/proxy/speechify',
  voiceId: 'geffen_32',
  model: 'simba-3.2',
  audioFormat: 'mp3',
  loudnessNormalization: true,
});

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();

Model selection

simba-english — English-only model optimized for latency. Best default for English voice pipelines.
simba-multilingual — Supports multiple languages; pair with the language option or let Speechify auto-detect.
simba-3.0 / simba-3.2 — Newer generation models with improved naturalness. simba-3.2 is the latest.

Tips

Emotion, pitch, and speed are controlled with SSML <prosody> tags in the input text rather than config options — see the Speechify SSML docs.
SpeechifyTTS is REST-based, not streaming. The full audio Blob is returned after the API processes the entire input. For real-time streaming, consider DeepgramTTS.
The Speechify API returns audio base64-encoded; SpeechifyTTS decodes it into a playable Blob for you.
Use proxyUrl in production so your API key stays server-side — set speechifyApiKey in your proxy config.