A video transmission method, system and a device includes a sending device that generates and sends a source video stream, where the source video stream includes a plurality of video data packets, each video data packet includes discard indication information, a discard priority indicated by discard indication information included in a first video data packet is higher than a discard priority indicated by discard indication information included in a second video data packet, video data in a video frame whose frame type is a non-reference B frame is encapsulated into the first video data packet, and video data in a video frame whose frame type is an I frame, a P frame, or a reference B frame is encapsulated into the second video data packet.