So, I read in Popular Mechanics that Jim Cameron uses a camera mounted on a helmet to scan faces of the actors. Just one? He got guts. I'd have insisted on a minimum 3 cameras. Left-Center-Right. That way, I can use intersecting line algorithm to get 3D data from reference points, instead of interpolating 2D data into 3D performance. Reading the text, that must've been a painful experience, indeed.