These models are designed to demonstrate recognizable figures.
They are shaded and rigged with realtime rendering in mind, but of course can be adapted for software rendering as well.
The Mr. T model features additional shader layers to achieve realistic wrinkling on his forhead when his brow is furrowed, and wrinkling around the eyes and smile lines when he grins. The face rig is made with joints that are controlled with set-driven keys in the channels of a control object.
Here’s a look at a work-in-progress lipsync test.
A view of the face rig joints: