If you've ever used a video editing program, you might be familiar with the concept of "keyframes," which define the beginning and end of a particular segment. Seeing where you came from and where you're going, the computer literally guesses what's in between, and creates smooth animation as a result -- the very same technique that students at Texas A&M University use to create motion capture that doesn't require arrays of cameras
or ping-pong balls
. Dropping the laws of Newtonian physics into their algorithms, Xiaolin Wei and Jinxiang Chai claim to have whipped up a computer program that can turn most any 2D video into simple 3D animation in real time, with just a few keyframes to start out. For instance, in a complex weightlifting segment 310 frames long where the camera panned, tilted and zoomed, animators had only to position eleven joints in thirteen keyframes (and make seven minute adjustments) to get the entire animation to turn out. See it in action after the break, or read their entire SIGGRAPH paper at our more coverage link.