The latest release of idiap/coqui-ai-tts is v0.28.0. This repository is a deep learning toolkit that synthesizes speech from text, the successor to Coqui TTS maintained by Idiap. The repository description presents the library as a tool validated in both research and production use. It was created on October 31, 2023 and currently holds 2338 stars.
The toolkit brings many TTS (text-to-speech), vocoder (a model that generates speech waveforms), and voice conversion models together in one place. It goes beyond using pretrained models as they are: it also carries tools for training new models or fine-tuning existing ones, plus utilities for analyzing and organizing datasets, so the flow of refining speech data, synthesizing it, and fixing it can continue inside a single library. It supports multi-speaker and multilingual synthesis as well as voice conversion, and offers both a command line and a Python API. The implementation language is Python, and the license is MPL-2.0.
What changed in v0.28.0 cannot be confirmed here, because the release notes are not part of this material. There are also no benchmark figures comparing performance in the published material. The momentum indicator recorded in the material is 1.0 out of 10, which is low. It is therefore hard to call this repository the current flagship speech synthesis project. It is instead worth trying directly for multi-speaker and multilingual speech synthesis or voice conversion, precisely because it bundles multiple models with training, fine-tuning, and dataset curation tools into one package.