To celebrate the beginning of Pride Month, here’s Erika from Pokemon, made in Stable Diffusion. The implications of which way she goes are nothing but a massive coincidence (I don’t think mid-1990s Japanese developers were even aware of the implications, and don’t know how solidified the rainbow symbol was). But still: All-female gym, and she gives the Rainbow badge.
I’ll just say that I haven’t been the only one to make that headcanon connection. Happy Pride Month!
Delighted to say that I’ve managed to wrangle Stable Diffusion enough to make some AI pseudo-photographs of Cholpon Murad-Kyzy, the heroine of All Union. Though not exactly matching my mind (what can be?) they’re still pretty close.
One of the most beautiful things about AI art generators is that it’s given me the change to take characters that I could see in my mind and bring them to life. With this in mind, I was especially delighted to create one of the main characters in The Sure Bet King. Behold Ly Rachany, CEO of Parilor, and lover of bunkers and security systems.
This weird Stable Diffusion picture was fun to make.
Basically, the prompt was something like “Currier and Ives style illustration of a woman in improvised metal armor”, followed by a lot of posing and inpainting. The result is this lady with a Union Civil War hat, a metal face shield, and an anachronistic weapon.
Some Stable Diffusion pseudo-photographs of the Undertale main humans. Chara is the black and white one with the single stripe, Frisk the color one with the two stripes. (The black and white one is partially because I made it in inspiration from a fanfic where Chara fell in 1926, so I chose one that fit the time period.)
One of the things I like about AI art is the ability to translate my words and mental images into pictures. This is a recent Stable Diffusion “portrait” of Amin Hayatov, president of the Union of Soviet Sovereign Republics in All Union’s present. The original picture had facial hair and something in his pocket that was inpainted out, and the Sovereign Union flag lapel pin was added in externally and then smoothed out through mild AI runs.
Had too much fun making this “recon aircraft photo” of wrecked armored vehicles in a field in Stable Diffusion. I actually found it was easier to just make the field and then inpaint in the wrecks. And if it’s a high altitude, somewhat blurry photo, that makes the imprecision regarding tanks and the like less important.
A common sight in September 1998 and after. Soviet air supremacy led to fields full of destroyed and abandoned Romanian vehicles throughout the country.
I just found this great article called “When Movie Artwork Was Great.” Long story short, the cover was an important part of advertising, and thus was worth the expense even for otherwise low-budget productions. A combination of computer tools making it easy to do a basic cover (Ie, it’s not like “you have to hand-draw it anyway…), and lower production values as the industry got squeezed meant that generic photoshopped mush took over.
It felt very familiar to me because the exact same thing happened concerning book covers for exactly the same reasons. Weirdly enough, otherwise high-profile books tended to have minimalist covers from the get-go. But trashy pulps that couldn’t even keep the main character’s name consistent (I’m not joking) would have spectacular covers from the likes of Gil Cohen and Ken Barr.
Now it’s just-look at the book section of a grocery store and you’ll know.
I mentioned the ability to improve book covers at very little cost as one of the upsides of AI art. Now that I’ve gotten more into it, I can say even more comfortably that something quickly makeable with the least controversial models (closed source, public domain only, etc…) and only small amounts of manual tweaking could leave most contemporary covers in its dust.
So yeah, shed a tear for the running silhouettes and clunkily shopped-in muscle men. I know we’ll all miss them so much.
My first AI art project is done. Variations on a theme. All were based on one image from somewhere in the mix and then changed via varying prompts (although the base image and seed were kept the same for all of them).
Enjoy “One Girl Through Countless Lenses: Vol. 1”
So I’ve fallen to the dark side and have begun making prompt-generated pieces of AI art. This is a very controversial subject with a lot of undeniably talented artists who I respect being furious about it, and understandably so. If I had to sum up my opinion on the controversy (beyond specific technical issues like how to treat stuff like training images for the sake of copyright and licensing), it’d be condensed to this:
But as a hobby, since I can write much better than I can draw, AI prompt tools have let me explore visual media in a delightful way. Yet what struck me when I really started getting into was how natural it seemed to me. And then it occurred to me: I’d done something similar before. Many, many times before. In wargames and simulators like Command, Nuclear War Simulator, Title Bout Boxing, and WMMA, I’d enjoyed simply creating a situation, allowing the RNG to add the needed element of chance to it, and then witnessing the result. And yes, frequently getting inspired by the result.
AI prompt tools allow me to do something similar with art and pictures. Yes, it can be an end. But a casualty list after a wargame scenario or results screen after a sports simulation can also be the beginning of a very human story.
As for AI writing, which is a thing, I’m strangely unfazed by it. I’m an artisanal sculptor, so seeing the metal casting factory rev up means little to my specific work. If that makes sense. Also, I’ve had the warped perspective of reading so many bad and mediocre books that I’m sincerely convinced that a computer can’t really do much worse.
Fuldapocalypse Fiction 