Facebook has thousands of pictures of our faces under a variety of lighting conditions. It would be exciting (or scary, depending on your point of view) to see that data, combined with sensors on the headset itself, to generate realistic-looking face models in the virtual space.
I just came here to post this. They also recently hired Yann LeCun to head up their AI Lab and also published a ground-breaking face recognition paper. So, they have the data, and they have the people.
