Humans can’t draw circles, not really, not perfect, anyway, without a heck of a lot of practice in extremely specific conditions. Computers are great at drawing circles–perfection is guaranteed–but they could never create an intricate watercolor New Yorker cover. Computers can’t sketch or improvise, can they?
“When people think of computer-generated images, they associate very distinct aesthetics with it. Mostly something very clean and/or complex, that looks way too perfect to be handmade,” Dörfelt explains. “These aesthetics oftentimes also come with a loss of individual artistic touch, because a lot of the personal gesture of actually drawing or painting something is delegated to the machine.”
But Dörfelt argues that the computer has no predefined aesthetics–not really. Everything you see has been programmed by a human. So in his programming, Dörfelt created general parameters for what noses, heads, and mouths looked like, then the computer created faces within these parameters. Think of the method as an algorithmic Mr. Potato Head: Rather than just mixing and matching parts from a library, every facial feature is dynamically drawn for every face.
Looking through all of Dörfelt’s cartoon people, you’ll spot similarity but not in the way that every PowerPoint presentation looks fundamentally the same. It’s more the similarity you sense when the same cartoonist draws multiple characters, or the same painter creates two landscapes. It’s the similarity that we’d likely call “voice” or “style.” Ultimately, Weird Faces Study is proof that CGI doesn’t need to look like CGI–assuming a programmer is clever enough to think outside the TRON-like grid of razor-sharp vector graphics–and instead places a bit of individual creativity at the core.
[Hat tip: Creative Applications]