I disagree - I thinks it will work on footage as well as on a still. I saw commercials with talking animals done in Flame a few years ago using similar tools with what AE has now ( Mesh Warp, FreeForm, Tracking- Mocha). I think AE has the necessary tools for the job.
I was referring to CrazyTalk software working with footage, which it doesn't. I haven't tried to do this sort of project in AE, which I'm sure it's capable of, although with a lot more work than in CrazyTalk.
With CrazyTalk, you import the still, choose the mouth and eye points, tweak them a little to create a 3D mesh, import the voice clip, and it does the facial animation automatically. The only limitation is that it has to be a still, and it pretty much can't go past a three-quarter profile, or the mesh looks odd. It will also add all sorts of mannerisms, such as head nods, blinks, and other things, which you can customize and keyframe.