Our AI overlords could employ an army of freakish chimera. At least, that’s the vibe I’m getting from Google’s new “Chimera Painter,” a web-based machine learning model that turns half-baked doodles (or in my case, beautiful sketches) into disturbing monsters.
Why would Google do such a thing? According to the company’s AI blog (which is written by humans), a tool like Chimera Painter could reduce the time needed to make high-quality art or game assets “without sacrificing artistic choices.”
To prove the potential for an AI monster-generator, Google’s Stadia team crafted a monster card game with creatures exported from Chimera Painter. It’s a compelling experiment that shows how AI could help speed up creative projects or game design. Tools like Chimera Painter could also inspire visual artists during a “sketching” stage or produce stand-in assets for game developers.
But how does Chimera Painter turn doodles into monsters? Google trained its AI to understand a 2D “creature outline” where different colors represent a monster’s head, legs, back, teeth, and other body parts. Interestingly, most of this training took place in Unreal Engine with 3D models. After designing a cool model and applying fantasy lighting effects, Google cut the model into segments so its AI could understand how each body part works together to create a fully-fleshed monster.
The trained Chimera Painter demo is free to use, and it could be a whole lot of fun for kids. That said, it’s hard to use the in-browser design tools, which is why Google allows you to “Load” illustrations made in programs like Photoshop. If you’re interested in AI image generators, you should also take NVIDIA’s landscape GAN or IBM’s AI-powered building designer for a spin.
Source: Google via TechCrunch