TorToiSe

TorToiSe is an open source text to speech program built around two priorities: strong multi voice capability and highly realistic prosody and intonation. The repository contains all the code needed to run it in inference mode, and it can be installed with pip from PyPI or directly from the GitHub repository. The model combines an autoregressive decoder with a diffusion decoder, which the author describes as the reason for its quality and also its slow generation. The README notes that early versions took around two minutes per medium sized sentence on an older GPU, while current releases reach a real time factor of roughly 0.25 to 0.3 on 4 GB of video memory, with streaming latency under 500 milliseconds. A live demo is hosted on Hugging Face Spaces, and the repository includes example outputs, a voice customization guide, an advanced usage guide, a Dockerfile and a Jupyter notebook. It is aimed at developers, researchers and hobbyists who want to synthesize speech locally and customize voices. A manuscript describing the method is linked from the repository. The project is released under the Apache 2.0 license and is free to use. There are no paid plans, and running it requires your own GPU or a hosted notebook.