Likewise I'm not a pro but this is what i know about it: Blendshapes are a good way to start - can make very controlled animations etc. After digging around and asking questions I've found a lot of studios use a combination of blend shapes with bones/clusters - Have blendshapes for the basic phenomes, and then use bones/clusters attached to SDK's/expressions for further detail and expression where needed. Gets around haveing a canned look which alot of pure blend shape animation seems to have. If you have the money both A|W and Gnomon have some videos of facial rigging which I'm sure will cover these methods in depth.