Disclosed is an AI voice interaction device, including a main body and an earphone detachably connected with the main body. User's voice is picked up by a voice pick-up, digital voice signals are obtained through analog-to-digital conversion, and are further transmitted to a first communication module through an earphone communication module. A body processor implements communication connection with a cloud server, and transmits to the cloud server for voice recognition and semantic analysis. Afterwards, the cloud server implements the functions such as making phone calls, sending short messages through the cellular transceiver, or calls corresponding network data, and transmits the network data back to the wristband-type AI voice interaction device to perform corresponding voice broadcasting according to the network data. Further, a TWS earphone and a wristband-type AI voice interaction system are disclosed.