Hmm, aside from all the other neat possibilities, it seems like that aspect would make it a lot easier to use representative video as template for generated content. E.g. you could record a sampler of positions for a video game avatar, and much more naturally composite them into movement over various terrain than is currently feasible.