Generate tracking points on a video to be able to composite screnes into videos or characters into virtual scenes.
Could be real-time, baked data, or both? Baked = analysis of video file for control point generation. Very cool for placing a video file into scene for augmented reality work. For example if a film scene had a mirror in it, then Vuo could be told the square control points for the mirror - then place a 3D quad with a real-time video camera into the movie mirror. By baking the data into the scene and using it for the 3D quad translation it would be very smooth and accurate.