The lip-sync is so good because apparently, they're using some sort of facial recognition technology where the actors face is digitally rendered by filming it with a bunch of special cameras. Basically, they create an exact digital copy of an actors face while they are performing the dialogue and copy it into the game. Seems a lot better than animating hundreds of different facial ticks and expressions, especially considering how dialogue heavy this game is supposed to be.
I heard it's got a 2000 page script.