Many everyday matching tasks taught to persons with developmental disabilities are visual-visual non-identity matching (VVNM) tasks, such as matching the printed word DOG to a picture of a dog, or matching a sock to a shoe. Research has shown that, for participants who have failed a VVNM prototype task, it is very difficult to teach them various VVNM training tasks using standard prompting and reinforcement. A potential rapid training procedure for teaching VVNM tasks would include withing-stimulus prompt fading. Such a training procedure, however, requires suitable teaching materials. In this paper we describe a strategy for creating computer generated fading steps, and illustrate their application for rapid teaching of VVNM discriminations. (Contains 1 table and 6 figures.)