Speechify plugin for LiveKit Agents (Node.js)

Speechify text-to-speech for LiveKit Agents on Node.js, maintained by Speechify.

This repository is the maintenance source for the plugin. Distribution is handled by LiveKit.

The plugin ships inside the livekit/agents-js monorepo and is published to npm by LiveKit as @livekit/agents-plugin-speechify. This repo mirrors that code so Speechify can maintain it, triage issues, and propose changes upstream. Bugs and contributions specific to the Speechify plugin are welcome here; releases are cut by LiveKit.

Installation

npm install @livekit/agents-plugin-speechify
# or
pnpm add @livekit/agents-plugin-speechify
# or
yarn add @livekit/agents-plugin-speechify

Authentication

Set your Speechify API key via the environment:

export SPEECHIFY_API_KEY="your-api-key"

Or pass it directly with apiKey.

Usage

import { AgentSession } from '@livekit/agents';
import * as speechify from '@livekit/agents-plugin-speechify';

const session = new AgentSession({
  tts: new speechify.TTS({
    voiceId: 'dominic_32',
    model: 'simba-3.2',
  }),
});

Options

Option	Default	Description
voiceId	'dominic_32'	Voice to synthesize with (see the Speechify /v1/voices endpoint).
model	provider default	simba-english, simba-multilingual, simba-3.0, or simba-3.2 (default).
language	provider default	BCP-47 code of the input, e.g. en-US.
loudnessNormalization	provider default	Normalize output loudness.
textNormalization	provider default	Expand numbers/dates into words before synthesis.
tokenizer	basic sentence tokenizer	Sentence tokenizer used to chunk input in stream().
apiKey	$SPEECHIFY_API_KEY	Speechify API key.
baseUrl	SDK default	Override the API base URL.
client	—	Pass a preconfigured SpeechifyClient from @speechify/api.

How it works

Built on the official @speechify/api SDK. stream() splits input into sentences and issues one /audio/speech request per sentence, emitting audio and aligned word-level timestamps as each sentence completes — near-streaming time-to-first-audio plus word marks (streaming and alignedTranscript capabilities). Audio is raw 16-bit little-endian PCM at 24 kHz mono; simba-3.2 is the default and recommended for the lowest time-to-first-audio.

Run it locally

A smoke runner exercises both synthesis paths against the live API (no LiveKit room needed), writes WAV files, and prints word timestamps + time-to-first-audio:

export SPEECHIFY_API_KEY="your-api-key"
pnpm install
pnpm example "Hello from Speechify."

Maintainers

Maintained by Speechify. Published and distributed by LiveKit as part of livekit/agents-js.