So I was looking at the really excellent face setup by the orange team over here:
http://orange.blender.org/blog/licence-to-drive
Brilliant stuff. Here’s what I’m trying to figure out:
- It seems like the best way to get faces that are balanced is to model half a face with mirroring on.
- And the best way to get shapes that balance well once you’ve finished modelling is to also to do it mirrored. That is, making half the smile and having it mirrored makes a perfectly balanced smile.
- BUT, what if you want to do asymmetrical things, like right eyebrow up, left lip down, etc? For that, you have to have a full mesh and so you need to apply your mirror modifier.
- BUT, once you do that, you kill any existing keys you have.
So, the big question is, how did they do that? Were they just very, very careful and very, very good? That’s certainly a possibility–they do genius work. Or is there a way to have your cake and eat it, too? To model and create symmetrical keys with mirroring on, and then, somehow, get those keys to survive once you apply the mirror modifier so you can add asymmetrical keys?
Bonus question: Assuming you can get you symmetrical keys to survive, what’s the best way you can think of to split them into assymetrical keys? That is, making a smile into a left smile and right smile by, say, doing something with the vertex groups on one side of the face and then the other, applying the full smile to the right vertex group to only get the right smile shape. Something like that 
I guess the big question underneath all of this is–if it’s not possible now, how soon wil will something like copying vertex positions from one shape to another (or one mesh/object to another) be possible? I think I remember Ton writing that that’s coming up, but I don’t know when.
-Mike