A multi-video stream transmission method and a device, where an access point sets an initial sending time and a sending period of an intra frame (I frame) of each station (STA) associated with the access point, and allocates an I frame service period and a predictive frame (P frame) service period to each STA according to the initial sending time and the sending period of the I frame of each STA, where each STA exclusively occupies a channel in the I frame service period and the P frame service period such that I frame sending time of STAs are staggered, network load balance is achieved, and it is ensured that a latency of each STA meets a quality of service (QoS) requirement.