Current uncompressed video servers are capable of streaming multiple video clips back-to-back in such a way that they appear to be a single uninterrupted stream. This is a relatively simple process made possible by frame boundaries equally spaced with no interframe dependencies. With the adoption of MPEG-2 and DV digital television standards, the distribution of video in compressed format will become more common. This change is fueling the development of video servers capable of distributing compressed video in broadcast-ready format. The seamless concatenation and splicing of streams that has been taken for granted in the uncompressed domain becomes complex in the compressed domain due to the mechanics of video encoding. This paper describes the problems associated with concatenating MPEG-2 transport streams (TS) and a technique to perform frame accurate seamless splicing from one MPEG-2 TS to another on compressed stream video servers.