Due to the lack of QoS support, ensuring an acceptable application level QoS for the real-time delivery of multi-object multimedia presentations on the current Internet is very challenging. Currently, the only feasible solution is to artificially "equalize" the variable network delays. In this paper, we first investigate the application-level objective performance measures which reflect the user-perceived quality, and then propose an adaptive delay and synchronization control scheme making use of those metrics in real-time. The proposed scheme adopts a distributed timing model and can be configured with two inter-stream synchronization options based upon the application?s own characteristics. We take a wireline video conferencing application consisting of one audio stream and one video stream as an example. We detail the operation of the proposed scheme and investigate the resultant performance. The simulation results show that small synchronization phase distortions, low MDU loss percentages and low average end-to-end delays could be achieved simultaneously. We further discuss a more challenging issue raised by mobile wireless applications. A preliminery solution is proposed.