A lot of little weirdness going on. Shadows in that last image are turning into a head of hair. The gravestone is spatially inverted. Definite limitations there.

But mostly I wonder, are you going to get the same face consistently? A lot of the examples here are famous people, so maybe those, but I suspect if you feed it three different renders of the same character you are going to get three different characters out. That's limiting.