This experiment was done with the help of Max Woolf on Twitter. The data scientist decided to make his own Pokémon by feeding AI information about the current monsters. The generator splintered out all the perfect monsters thatPokemonfans can never see. Bonus batch! pic.twitter.com/qdr7SoesUj. Max Woolf (@minimaxir) December 15, 2021. “The AI used here is a fine-tuned RUN-E on the official Pokemon image (no VQGAN + CLIP) or Wombo Dream. Unfortunately it’s not so portable/easy to use otherwise I would make an interactive demo,” said Woolf on Reddit. “The way in which the AI works is that it collects the images from the top to the right in eight chunks. It samples the next chunk fairly randomly so the image is consistent. The finetuning process teaches the AI how to properly recognize a Pokemon. As you can see above, those monsters, who fans affectionately call Fakemon, are very good. It would be a lot of sense to do an artist, if not the same as the AI. This Pokemon, though absolutely foreign to us, have the signature style you’d expect from GameFreak. They’d translate so well into anime. Nevertheless, these Monkeys are little more than a test run. After Pokemon is run, there are plenty of monsters joining the roster. The company’s developers are a better player than anyone else. Their work shines in their work. But if they ever get the art off a note, you can always study this wild guy for a short time until inspiration hits me. What do you think of these wild new Pokemon? Did you consider the generator nailed the assignment or…?
