The released model is for academic purposes only. The main model is trained with Chinese and English audio data of 100,000+ hours. The open-source version on HuggingFace is a 40,000 hours pre-trained ...