A method and apparatus for synchronizing streaming media with multiple output devices. One or more media servers serve media streams to one or more output devices (i.e., players). For playback synchronization, one output device is the “master”, whereas the remaining output devices are “slaves”. More data is requested from the media server by the “master” device to maintain a nominal buffer fill level over time. The “slave” devices receive streamed data from the media server at the rate determined by the master device's data requests, and the average rate of data flow over the streaming network is thus controlled by the frequency of the single “master” device's crystal. “Slave” devices make playback rate corrections to maintain respective buffer fill levels within upper and lower threshold levels. For slow networks, each media data packet timestamp is calculated from the time the master's buffer reaches nominal level.