ChatTTS is a text-to-speech model specially crafted for conversational scenarios like LLM assistant. It provides conversational TTS tailored for interactive dialogues involving multiple speakers, ensuring a natural and expressive speech synthesis. The model stands out in predicting and managing detailed prosodic elements such as laughter, pauses, and interjections, outperforming many open-source TTS models in prosody aspects. Leveraging a core model trained on 100,000+ hours of Chinese and English audio data, ChatTTS facilitates further research and development through pretrained models. The platform's future plans encompass releasing base models as open source, enabling streaming audio generation, and introducing versions with multi-emotion control. It is crucial to highlight that ChatTTS is meant for academic and research purposes exclusively, and users are advised to use the technology responsibly and ethically. For queries regarding the model and future plans, users can reach out to the team at
[email protected].
