Abstract: Todays networks are characterized by heterogeneous environments. Various transmission techniques coexist and offer different bandwidths at different costs. Because of videoconferences and similar applications becoming more and more important this paper presents the architecture of the video communication system "XNetvideo" which is scalable in terms of SNR (signal to noise ratio)-scaling, image size and frame rate. Xnetvideo is a tool for video transmission over IP-networks and especially ATM, that was developed at our Institute. The description of this architecture, and the discussion of implementation issues and performance results are the major topics of this contribution. Moreover a brief survey of existing tools and architectures in this area is given.