Hey GDLtd -
So, since posting this I've been working with it a lot, so here are some of my thoughts:
1) It's genius.
2) The name says it all : ADAPTABLE facial setup, and boy is it adaptable. Everything is designed for you to plug your characters into it. I'd say if your character has good clean topology (and is symmetrical!) you can plug it in in a day or two (day 1 is reading the docs and making mistakes).
3) The UI is brilliant, and you'll find that the Corrective Blendshape System allows you to go the extra mile and really tweak the results down to the vertex level.
4) Non-linearity is in fact a huge improvement over standard blendshapes. In fact I can't even look at standard blendshapes anymore.
Seriously, it makes a difference that is difficult to describe but impossible to miss: think about it: ALL the points on the face are now following mocaped, natural curves and paths, instead of straight lines. I can honestly say, I will never work with linear blendshapes again. since trying this tool I've thrown out my 3 months of work of building blendshapes by hand which, now that I look back at it, hurts my eyes.
5) The movements are based on FACS, in fact they're not just based on FACS, they're mocaped clips of FACS AU's. So, for a Jaw_open, or a Lip_Raiser, you'll get all the extra tapering on the cheeks for free, and you won't have to sit there and pull points and stare at a mirror for weeks on end.
Now for the cons - there are only three as far as I can tell:
1) it is not quite as fast as blendshape nodes. If you run a scene with just your character's head in it, it's perfectly fast and a joy to work with. But I'm not sure how it responds if you drop your character in a heavier scene.
2) Because each motion clip is mocaped, it starts and ends with a tapering effect, like a spline curve flattening out. This can be a drag to work with, especially if you want to plug certain expressions on opposite ends of one control Osipa-style. There is the option to crank down the tapering with an envelope, but then you lose the non-linearity... There is a solution, read on.
3) There is, so far as I can tell, no way to sculpt corrections on combinations between expressions. If you do a Jaw_Open + Lip_Funneler, for instance, you'll find the lips funnel out too far, and there's no way to go in and tweak that. Yes, you can do it by hand, by doing an extra blendshape driven by expressions, but THAT'll be linear, and you have to plug it in all by hand. It's fine, but I have a preferred solution, read on.
Because of the shortcomings above, I've come up with a slightly different workflow. I basically pair up AFS with Daniel-Pook-Kolb's BCS system. (see Spiraloid, or http://www.tokeru.com/twiki/bin/view/Main/MayaRigging). What this means is that once I've done the whole AFS homework up to being animation-ready, I then pipe out blendshapes from this setup, trash the whole AFS interface, and plug your blendshapes into BCS. But you make, say, 2 or 3 blendshapes PER expressions (30%, 60%, 100%) and plug them in with arc interpolation into BCS, so it's NOT linear. The points keep their paths, albeit paths with lower sampling resolution, which is exactly what is needed to get rid of the tapering effect.
The beauty of this is that it preserves, I'd say, 90% of the quality of AFS, but then gives you BCS's incredible flexibility for sculpting combinations.
Essentially, I'd say those two plugins pair up very powerfully. But if you only want to work with AFS, I think you'll be very pleased with the results. It gives you that extra, organic quality of real captured motion.
Hats off to the the guys at Institute of Animation for a really well-rounded, user-friendly, forward-thinking new approach to facial animation! 