I’ve gone back and forth on AI art that’s cool and scary, but I can’t deny that these fusion renders of Overwatch heroes (opens in a new tab) I saw on Reddit last week they are pretty sweet. It’s fascinating that we can now ask an AI for an opinion on what Soldier 76 and Mei would look like together, and it can spit out some surprisingly cool and possibly original character design.
That’s what author Drew Harrison did when he became one of the few thousand with access to Stable Diffusion. (opens in a new tab), an AI image model similar to other popular tools like DALL-E and Midjourney. Harrison started by feeding the AI simple instructions like “make a fusion portrait of Mercy from Overwatch and Junkrat from Overwatch,” and in some cases, took more direct control over artistic influences with tags like 4K, highly detailed cinematic lighting. Having played around with the DALL-E Mini online tool a bit, he didn’t know you can be so specific with instructions.
The results are surprising:
Image 1 of 5
I’m particularly intrigued/puzzled by the Mercy x Junkrat (Mercrat) glow and the arm-mounted belly on Moira x Roadhog (Moidhog) arm. You can’t often expect an AI to spit out a fridge-worthy sketch on its first try. Harrison told me that the art of AI is often a trial and error process.
“Try a message, see what comes back, check, run again and repeat until you get the image you want,” Harrison said. Speed is why Harrison currently prefers Stable Diffusion over other AI art tools. He describes Stable Diffusion as “lightning fast and light”, two big advantages when the only creative bottleneck is render time and VRAM.
Stable Diffusion is also the biggest AI art tool in the process of becoming open source, though it sounds like you’ll need a fairly robust platform. (opens in a new tab) to run it
However, the art of AI still has some clear limitations. As you can see from Merccrat’s weird octo-hand, the AI models still don’t seem to know what to do with their fingers and palms. Sometimes there are too many fingers, or not enough, or just a fuzzy mess. The AI also tends to get weirder in proportions when it mimics humans, though Moidhog is an extreme example. Check out the full gallery of Overwatch melds here (opens in a new tab).
While I hope Mei 76 will one day become a true Overwatch hero, the rapid proliferation of AI art (opens in a new tab) makes me a little uneasy. Stable Diffusion, like any other AI, is trained on a dataset of internet images pulled from everywhere and everywhere. As Harrison showed me with “a fusion of Joe Biden and Barack Obama painted by Van Gogh, (opens in a new tab)“These tools are capable of searching and imitating the style of a specific person.
One could argue that this is no different than a person being inspired by another artist, but when we’re talking about a computer brain storing millions of copies of other people’s art, it’s hard to tell where inspiration ends and copying begins.