A method for stream synchronization of real-time multimedia transport over a packet network, in which the multimedia includes a first stream and a second stream which are buffered and played through buffers. The method includes: measuring average transport delays of the first and second streams; measuring transport delay jitters of the first and second streams; and calculating a delay difference between the first and second streams which corresponds to the average transport delays and the transport delay jitters of the first and second streams, and setting buffer durations. A device for stream synchronization of real-time multimedia transport over a packet network is also provided.