clone-voice

Clone any voice from a short audio clip and generate speech.

Quick start

import clone from 'clone-voice';

const voice = await clone('./my-voice.wav');
const wav = await voice.speak('Hello from my cloned voice!');

clone() loads models, encodes the voice, returns a Voice. voice.speak() returns a WAV ArrayBuffer. That's the whole API.

From mic

import clone from 'clone-voice';

const voice = await clone.mic(5000);
const wav = await voice.speak('Cloned from my mic!');

Works in browser (MediaRecorder) and Node/CLI (sox).

Built-in voices

const voice = await clone('alba');
const wav = await voice.speak('Using a built-in voice.');

Streaming

const voice = await clone('./my-voice.wav');

voice.on('chunk', (pcm) => player.play(pcm));
await voice.stream('Streaming speech.');

Node.js / CLI

import { writeFile } from 'node:fs/promises';
import clone from 'clone-voice';

const voice = await clone('./voice.wav');
const wav = await voice.speak('Hello from Node.');
await writeFile('./output.wav', Buffer.from(wav));

Multi-language

const voice = await clone('./stimme.wav', { lang: 'german' });
const wav = await voice.speak('Hallo, wie geht es Ihnen?');

Supported: english_2026-04, german, italian, portuguese, spanish.

Credits

Pocket TTS by Kyutai
ONNX Web models by KevinAHM