What, Face-Over Actors Now?

The New York Times has an article about Image Metrics, a company over in Santa Monica that's taking a computer vision approach to capturing facial motion and applying it to 3-D character faces.

The results aren't bad. I wonder how much computing power is needed and how long the process takes. Also wondering if there's any cleaning work, and how advanced the facial rigs (in Maya) are for these examples. It's clear from the video that this is not a real-time technique (yet, anyway). It only works with a static camera, and the face and body have to be captured separately. I would bet these factors diminish the coherency of the performance. Might be useful for smaller scale projects though.

Thanks to Slashdot.