The present invention provides a low-latency streaming HLS implementation method, the method includes: dynamically generating an independent m3u8 list for each playback request, dynamically and quickly generating a small slice only for this playback request from the media data cached on the server File, in this way the traditional HLS stream processing method. When the user makes a playback request to the edge server, a UUID is automatically assigned to the user as an identification. The edge server starts RTMP back-to-source streaming, quickly starts slicing the RTMP stream data to be played, and generates an independent m3u8 list. The player gets The list and the first 3 slices are ready to play. The present invention is exactly the same as the RTMP stream when the stream is returned to the source, and there is no need to build an HLS slicing system. The stream media and RTMP use the same cluster, which can save a set of equipment groups, N computer rooms and equipment. This method is simple for customers to configure, easy to use, fast to go online, simple domain name CNAME, and can reduce latency. Compared with the 40s latency of traditional HLS, HLS+ can be at the same level as RTMP.