Facial rigging workflow

- Must be non-destructive. 
- Games or production? Speed vs Quality? Predictable or Experimental?

Games:
Joint based. That would mean scaling is something we should take note about. 

Production:
Shapes fidelity is priority. That means stacked/layered blendshapes for great shapes.
1. joint based simple rig. Jaw, teeth, tongue, head, eye balls.
2. Blendshapes expressions. All lips, all cheeks, all eyelids, all eyebrows etc.
3. Non-linear deformers add-on. Lattice, squash/stretch.
4. preBindMatrix. <-- important. This is how to tweak with onMesh controllers.

Speed:
Use auto rigging solution like Advanced Skeleton.

Quality:
Worth throwing resources for some RnDs. 

Predictable:
Use auto rigging solution like Advanced Skeleton.

Experimental:
Anything goes. RnDs.