Supertonic Zh

Supertonic Zh is a Mandarin Chinese text-to-speech model available on Hugging Face. It is designed to generate speech audio from input text in Mandarin Chinese, and supports code-switching. The model is implemented in PyTorch and is compatible with the Supertonic library, which allows users to synthesize speech by specifying voice styles and saving the generated audio. Example usage involves initializing the TTS model, selecting a voice style, synthesizing speech from a given text, and saving the output as an audio file. The model weights for Supertonic Zh are distributed under an evaluation license that restricts usage to non-commercial purposes. This limitation is due to the inclusion of the Baker/CSMSC corpus in the training data, which itself is licensed for non-commercial use only. Access to the model also requires agreement to the BigScience Open RAIL-M use-based restrictions that apply to the base Supertonic-3 model. The evidence does not specify further details about the range of voice styles, supported features beyond code-switching, or additional technical specifications. Supertonic Zh is intended for tasks involving Mandarin Chinese text-to-speech synthesis, and is suitable for research and evaluation in non-commercial settings. The model can be used with various platforms and tools that support Hugging Face models and the Supertonic library, including local applications and notebooks such as Google Colab and Kaggle. No information is provided about commercial licensing, additional integrations, or extended capabilities beyond those described.