Chatterbox
Resemble AI's MIT-licensed 0.5B voice-cloning TTS, Multilingual V3 with 6 language packs. Retired in Aug 2026: archived, no updates.
In short. Chatterbox reads text aloud and can copy a voice from a short sample. It runs on one GPU, covers about two dozen languages, adds an inaudible watermark to every clip and uses the MIT license. Its maker stopped selling and maintaining it in August 2026, so it is now a frozen archive.
Summary
Chatterbox is a family of open text-to-speech models from Resemble AI. The original English model came out in 2025, and a 23-language version followed later that year. Chatterbox Multilingual V3 was released on 10 June 2026 with the same 0.5B Llama backbone. Resemble says V3 improves speaker similarity, reduces off-prompt output and repetition, and sounds more natural. Its announcement lists 21 languages plus 4 regional variants, and six dedicated single-language fine-tunes cover Chinese, Hindi, Brazilian and European Portuguese, and Latin American and European Spanish.
Chatterbox clones a voice from a short audio prompt, and it has an exaggeration setting and a CFG weight to control how dramatic the delivery is. Every output carries a PerTh watermark. The V3 checkpoint files total about 3.2 GB, and pip install chatterbox-tts installs the code. Hugging Face counted about 2.4 million downloads over the 30 days to 11 November 2026.
The key caveat is its status. In August 2026 Resemble AI said it had moved to deepfake detection and no longer sells voice generation. The GitHub repository now says Chatterbox is retired research and is no longer maintained: the code stays public under the MIT license as an archive, with no support or updates. It still works and the license is permissive, but expect no fixes or new languages.
Good for
- Voice cloning and expressive narration from a short reference clip, with an exaggeration control
- Multilingual voice agents and localization work under the MIT license
- Teams that want watermarked output by default
Watch out for
- Resemble AI retired Chatterbox in August 2026, so the repository is an unmaintained archive
- The model card lists 23 languages while the V3 announcement lists 21 plus 4 variants, so test your language before committing
- Cross-language cloning keeps the accent of the reference clip, and the clip should match the language tag
- Outputs carry a PerTh watermark
Performance claims come from the developer and have not been independently verified.
How much memory does it need?
Memory figures for each variant of Chatterbox. Image, video, speech and music pipelines also load a text encoder and a decoder, so the memory needed can be larger than the main checkpoint.
| Variants | Parameters | Download size | Memory needed | Fits on |
|---|---|---|---|---|
| Chatterbox Multilingual V3 | 500M | 3.21 GB | 3.69 GB | Laptop or phone |
Download size covers every component needed to run the model, such as the generator, text encoder and decoder. Memory needed is a stated or published measured minimum where one exists; otherwise it is the download size plus about 15%. Larger images, longer clips and higher resolutions need more.
The license in plain words
The shortest and simplest permissive license. You can do almost anything with the model as long as you keep the copyright and license notice.
See also
- Best open-source text-to-speech models in 2026
- Open speech models: transcription is open, text-to-speech is mixed
Frequently asked questions
- What is Chatterbox?
- Chatterbox reads text aloud and can copy a voice from a short sample. It runs on one GPU, covers about two dozen languages, adds an inaudible watermark to every clip and uses the MIT license. Its maker stopped selling and maintaining it in August 2026, so it is now a frozen archive.
- Can I use Chatterbox commercially?
- Yes. Chatterbox is released under MIT, a permissive license. You can use it commercially if you follow the standard conditions, such as keeping the license notice.
- Is Chatterbox open source?
- Chatterbox is released under MIT, a license approved by the Open Source Initiative. Training data and training code may not be public, so it is best described as open weights under an open-source license.
- Can I run Chatterbox on my own computer?
- Yes. The smallest variant, Chatterbox Multilingual V3, needs about 3.69 GB of memory at 4-bit precision, which fits a modern laptop or a mid-range graphics card.
- When was Chatterbox released?
- Chatterbox was released by Resemble AI on June 10, 2026.
Official links
Sources
Specifications come from these primary sources. Benchmark claims, where mentioned, are reported by the developer.
- Chatterbox Multilingual v3 announcement (Resemble AI, 10 Jun 2026) resemble.ai
- Chatterbox repository, with the August 2026 retirement notice (GitHub) github.com
- Chatterbox model card (Hugging Face) huggingface.co
- Why Resemble AI went all in on deepfake detection (13 Aug 2026) resemble.ai
Release date and facts last verified November 11, 2026.
Related models
| Model | Released | Parameters | Context | License | Smallest variant needs |
|---|---|---|---|---|---|
| Kokoro hexgrad | 82M | – | Apache 2.0 (Permissive) | 0.38 GB Laptop or phone | |
| Pocket TTS Kyutai | 18M – 100M | – | CC BY 4.0 (Permissive) | 0.127 GB Laptop or phone | |
| VoxCPM2 OpenBMB | 2.29B | – | Apache 2.0 (Permissive) | 8 GB Laptop or phone | |
| Qwen3-TTS Alibaba Qwen | 910M – 1.92B | – | Apache 2.0 (Permissive) | 3.68 GB Laptop or phone |