“Unlike every other motion capture thing you’ve ever seen, this is a full performance capture system. It doesn’t just track movement, or grab animation data from actor’s faces as they speak their lines, it captures everything about an actor’s performance, and generates a fully-textured 3D model based on what it sees and hears.”
“It uses 32 high definition cameras (divided into 16 stereoscopic pairs) to capture every angle of an actor’s performance at 30 frames per second. From this data it generates a fully-textured 3D model (at the moment it’s just their heads, but later it will be full body) that incorporates every nuance, mannerism, and emotional detail from the performance.”
And the fully-textured 3D model is apparently run on-the-fly, in real-time.