XTTS v2

Voice cloning from a few seconds of audio, in seventeen languages, running locally. The company behind it shut down; the weights remain.

At a glance

Made by
Coqui
Family
XTTS
Licence
Coqui Public Model License
Graphics memory
4 GB minimum, 6 GB recommended
Adult content
UnknownWe have not found a clear answer. We would rather say so.

The licence is not an open-source licence — it is a bespoke model licence that restricts commercial use. Coqui the company closed in January 2024, so nothing here is maintained by its authors any more, and the community forks are where the activity went.

It is the clearest consent problem in this catalog: it needs six seconds of someone's voice and asks nothing about whether you have their permission.