Bert-VITS2

Why we included this project

Bert-VITS2 gets its natural, expressive voice quality from pairing the VITS2 text-to-speech backbone with BERT embeddings. It trains on multiple languages, and the webui-based preprocessing and training workflow gives you fine control over voice characteristics, which makes it a good fit for teams that want to build custom voices, clone a speaking style, or drop speech synthesis into a larger product instead of paying for a closed commercial API. If you are comfortable fine-tuning PyTorch models, the code is approachable and clearly inherits from the open-source VITS and MassTTS lineage. One caveat worth knowing: the maintainers now point new users to their successor project Fish-Speech for ongoing development, so look there first if you are starting fresh.