An end station for a videoconference / multimedia conference is disclosed, where the end station requests, receives and displays multiple video streams. Call control messages request video streams with specified video policies. A static policy specifies a constant source video stream, e.g., a participant. A dynamic policy dynamically maps various source streams to a requested stream and shows, for example, the current speaker, or a round robin of participants. A network access device, e.g., a media switch or a video composition system, mediates between the multi-stream end station and the core conference system. Multi-stream endpoints need not handle the complexity of directly receiving video according to a potentially wide variety of call control protocols, formats, and bit-rates. Multi-stream endpoints decentralize compositing video streams, which increases functional flexibility and reduces the need for centralized equipment.