Chatterbox TTS
Open-source family of text-to-speech models with zero-shot voice cloning, multilingual output, and emotion exaggeration control.
At a Glance
About Chatterbox TTS
Chatterbox is a family of open-source text-to-speech models from Resemble AI, released under the MIT license. It generates speech from text, optionally cloning a voice from a short reference clip. The README states that as of August 2026 Chatterbox is retired research and no longer maintained, and the code remains public as an archive.
What It Is
Chatterbox is a Python library and set of models for zero-shot text-to-speech and voice cloning. Users install it with pip or from source, load a model, and call a generate function with text and an optional audio prompt to produce a waveform that can be saved as audio. Example scripts cover TTS, Turbo, Nano and voice conversion.
Model Family
The repository documents several models. Chatterbox-Turbo (350M parameters, English) targets low-latency voice agents and supports native paralinguistic tags such as [laugh], [cough] and [chuckle]. Chatterbox-Nano (110M) shares Turbo's architecture and targets on-device and CPU inference; the README says it runs 3x faster than realtime on 8 CPU cores. Chatterbox-Multilingual V3 (500M) covers 23 languages and aims to improve speaker similarity and reduce hallucinations. A Single Language Pack offers dedicated finetunes for Chinese, Latin American and Spain Spanish, Brazilian and Portugal Portuguese, and Hindi. The original English Chatterbox model exposes exaggeration and CFG weight controls.
Watermarking and Evaluation
Audio generated by Chatterbox includes Resemble AI's PerTh neural watermark, and the README provides a script to check audio for it. The README also links to Podonos evaluation reports comparing Chatterbox Turbo with other TTS systems; these are vendor-published comparisons.
Current Status
The README says Resemble AI no longer sells text-to-speech or voice cloning products and that this repository offers no support or updates. The latest tagged release listed is v0.1.2.
Community Discussions
Be the first to start a conversation about Chatterbox TTS
Share your experience with Chatterbox TTS, ask questions, or help others learn from your insights.
Pricing
Open Source (MIT)
Chatterbox TTS core code is released under the MIT license. The repository is retired research as of August 2026, with no support, updates, or paid tier.
- MIT License
- Self-hosted open-source TTS models
- Retired research, no longer maintained as of August 2026
- No support, updates, or paid tier
Capabilities
Key Features
- Zero-shot voice cloning from a reference audio clip
- Emotion exaggeration and CFG weight controls
- Paralinguistic tags such as [laugh], [cough], [chuckle] in Turbo and Nano
- Multilingual V3 supporting 23 languages
- Single Language Pack finetunes for six languages and dialects
- Chatterbox-Nano for CPU and on-device inference
- Built-in PerTh neural audio watermarking with extraction script
- MIT-licensed open-source code
