The general idea behind it, is to take a standard 3d character animation workflow and add in video and motion capture data to try and inject an organic feeling into the motion. 3d animation has a habit of looking dead and mechanical in it's motion, especially for facial movements.
So far so good. The characters are real bodies on greenscreen, they wear a DIY tracking harness in order to track and replace their heads with 3d heads. As a separate process, the voice artists are filmed with facial track points, which are used to drive the movement of the heads. Finally heads and bodies are combined with 3d environments, camera movements, 2d textures to create the end result.
One big advantage to this technique is that lip synching is automatic from the facial track points, though it did require some fairly tedious scripting in order to build in custom controls to the rig (sliders to scale the relative influence of each track point on the face mesh).
Here are some samples from the work in progress
