I was going to say that the overlaps between one lock of hair to another, between hair and background/hair and skin, and the edges of lips and teeth looked a little poorly keyed, like the objects had high resolution textures but the surface map/motion rig was low-poly...which would be annoying but probably easy enough to ignore in an indie film.

But live? On a single (hard to get) consumer GPU? That's seriously impressive. It makes me wonder how much of this is hand-tuned rigging and how much is physics based; if you tried to shake hands with this digital human using a game controller or VR rig, how would that look?