Because they're trying to make something that doesn't exist. Ai is really good at making stuff that already exists, or even merging traits from multiple things that already exist. It can't create stuff that doesn't exist, though.

So, what the creator does is point Stable Diffusion at a prompt that says something like "Jesse from Breaking Bad, but Russian", point a voice generator primed on Jesse's lines at what he wants Soviet Jesse to say, and point a third one that mimics face movements to the picture and the Ai-generated voice line.

Now he has a picture with a moving face that talks and says whatever dumb shit he wants it to, for minimal effort but all the internet points.