Moshi is a speech-text foundation model and full-duplex spoken dialogue framework. It uses Mimi, a state-of-the-art streaming neural audio codec.
By chatting or signing in you agree to the Terms and chat-message logging (revocable in History).