A system clock reference is transmitted to user devices over a short range communication channel and used to calculated a clock offset. The clock offset is stored together with a user identifier in a server memory, from which the clock offset can be retrieved when receiving the user identifier together with a video stream of video frames tagged with respective timestamps. The timestamps are converted into capture times according to a system clock using the retrieved clock offsets. Video streams from multiple user devices thereby have video frames timestamped using a same clock reference and can be time aligned.