Disclosed is a system for transferring real-time audio/video stream. The system comprises a mobile device, a server and a client terminal. The mobile device comprises a recording module, a fragmented media data (FMD) codec, a segmented media data (SMD) codec and a wireless communication module. The recording module records an event on the scene and generates a real-time video stream. The FMD codec encodes the real-time video stream into fragmented media data comprising a start fragment information, at least one media fragment and a media fragment (MF) random access point. The SMD codec directly segments the fragmented media data into a plurality of segmented media data, which are a start segment and at least one play segment. The SMD codec automatically and sequentially transmits the segmented media data to the server. Finally, the server pushes the segmented media data to the client terminal for play through a browser.