This project utilizes the advanced streaming neural audio encoder, Mimi, to implement a voice system capable of real-time, simultaneous listening and speaking (full-duplex), and natural dialogue. It not only understands what you say but can also answer your questions like a normal conversation, with almost no delay in the process and allowing interruption at any time during the dialogue. However, it currently only supports communication in English and can be used to practice spoken English, helping to overcome the barrier of speaking English because the interlocutor is just an AI.