Chatterbox

Model

Open-source Resemble AI models for voice conversion and text to speech

Price: Free, open source
Access: None, runs locally

About

MIT-licensed Python package and weights from Resemble AI. ChatterboxVC re-voices a recording to match a short target sample, zero-shot; the TTS models clone voices in 23 languages, Turkish included. Runs on CUDA, Apple MPS or CPU. Every output carries an inaudible PerTh watermark.

What you can do with it

Convert a speech recording into another voice from a short reference clip
Re-voice narration or dialogue on your own GPU without sending audio to a vendor
Generate speech in a cloned voice in 23 languages with the TTS models

Get started

Install Python 3.10 or newer, then pip install chatterbox-tts
Pass the source recording and a clip of the target voice

Example

import torchaudio as ta
from chatterbox.vc import ChatterboxVC

model = ChatterboxVC.from_pretrained("cuda")  # or "mps" / "cpu"
wav = model.generate(
    audio="source.wav",
    target_voice_path="target_voice.wav",
)
ta.save("converted.wav", wav, model.sr)

Details

Hosting: Runs locally
Available in: Worldwide
Official SDKs: Python
MCP server: None