Manipulating digital video using desktop computers, especially in uncompressed form, and communicating it over networks can be time-consuming because of it large bandwidth requirements. While the size of a videoframe is large, the amount of new information in each video frame is open quite small. Wepresent an algorithm that predicts those parts of the video image that change from frame to frame. This effectively allows the manipulation of uncompressed digital video using common personal computers with no special hardware support. The algorithm uses a statistical technique and takes advantage of locality of change. We show that the algorithm is effective and yet is simple and low-cost.