I realize that I'll probably never get something like this, but dang, that was cool.There seemed to be a distinct quality difference in the realtime created models, and the prerecorded models. The realtime ones had issues, where as the prerecorded ones (like the girl with the dog) were smooth and nice. I definitely think that if this goes mainstream, higher quality models will be generated in most cases and only use real time stuff if there isn't one available.