Chatterbox Multilingual V3 🧠 Model Open source

What it is

Chatterbox is Resemble AI's open-source text-to-speech family, released under the MIT licence. The current general-purpose model, Chatterbox Multilingual V3 (June 2026), is a 0.5B-parameter model that clones a voice from a short sample in 23+ languages. Next to it sit Chatterbox-Turbo (350M, English, low-latency, with paralinguistic tags) and Chatterbox-Nano (110M, English, runs on a CPU).

Key features

• Zero-shot voice cloning from a short reference clip
• 23+ languages, with dedicated packs for Chinese, Spanish, Portuguese and Hindi
• Turbo supports tags like [laugh] and [cough] and targets about 75 ms latency
• Emotion exaggeration and guidance controls
• Every output carries Resemble's PerTh watermark for detecting synthetic audio

Where you can use it

Install it with pip and run it locally, or use the Hugging Face demo. Resemble AI also offers Chatterbox through its own hosted platform, and NVIDIA offers it as a NIM microservice.

Pricing and rights

Free under MIT for commercial and personal use. You pay only for your own compute, or Resemble AI's hosted rates if you use their service. The built-in PerTh watermark makes outputs traceable. Because cloning is zero-shot, only clone voices you own or have written consent to use.

Who it is for

Developers and creators who want free, self-hosted voice cloning for narration, dubbing or characters, with provenance built in.

Verdict

Chatterbox is one of the strongest permissively licensed cloning models, and the watermark is a responsible default. Running it well locally still takes a GPU and some setup. Multilingual quality varies by language.