DOOM has been ported to Stable Diffusion. As in, this is literally being run on nothing but stable diffusion frames and a "memory" that allows the machine learning algorithm to infer what the next frame should look like based on input. It's literally making all of this up in real time.
@coded_artist @Hoss Combined technology. Level geometry and characters are made in the classic way. AI makes stuff that is more random like fog, falling leaves pebbles rolling down and whatnot.
Remember that this was trained on inputs and frames from the existing game. The game had to already exist to be modeled by the AI system. It didn't create a game from descriptions, it just made an incredibly inefficient copy of an existing one.
@gentoobro @Hoss this is indeed the situation... for now.
AI powered animation went from helmet wearing retard puppets bumping into walls, to a showcase of martial artists, in just several years.
For this application, AI is being underestimated in my opnion.
I think it will be easier at some point to draw a basic scene, I mean really basic shapes as instructions to an AI, not just lower resolution like we already do now, and have the AI do most if not all the rendering work.
