One of the cool things about the kinect is that the depth image is a really good source to use for edge detection / outline stuff, and it's easy to key out depth.

This is sort of a reinvention of an old technique of mine ("g-body.qtz"), but using the kinect as a source, instead of regular video, and adding a number of other elements that set it up to work better, and work in Snow Leopard as well (the queue inside of an iterator doesn't work the same in Snow Leopard in this instance, so a great deal had to be changed.)