Closer resemblence = better output?
I wonder if the output becomes more believable if the original source video subject is closer to the target face already. What if you hire some lookalike actor to act out the fake you want to make, would it look more convincing?