Since I started to experiment with AI generated art back in August 2022 a lot has gone down. Chief among the developments, a lawsuit filed by a group of artists who claim the Midjourney creation site, the StableDiffusion AI engine, and DeviantArt, has plagiarized their work. Meaning, the AI engine was trained on images pulled from the DeviantArt site, which is a showcase for fannish, SFF, and out-there art, without the artists’ permission, and used by Midjourney to create original art for its users. Which has opened up a whole can of worms considering that some of those users are now profiting off those works. Many copyright lawyers will be very busy in the future over this. History is being made.
And all this means, too, that AI art has since exploded into the public consciousness. I feel a little hipsterish right now, that something I’ve discovered and thought of as cool now has been taken up by everyone else, and the ethicality of it troubles me also even as I want to experiment more and see for myself what it can do and can’t do.
One of the things I’ve found is how Midjourney differs from all the other generators, even those using StableDiffusion themselves, and it’s not its access to the whole of the DeviantArt site. The images are too good, too polished, and leads me to believe something else is going behind the scenes. But I can’t put my finger on it. That it’s run off a Discord server is still fishy to me.
So I came up with a prompt inspired by a photo of two people that I know. I ran that photo through a Hugging Face site that generates a prompt from a visual input, and got something that said, basically, “Rockabilly couple, orange jumpsuit, glasses, suit, leopard skin.” Which amused me greatly as the couple in question were not Rockabilly and not wearing any orange jumpsuits; they were Native American. But there you have it.
I added gouache as the medium.
These are my favorites of the four pictures generated. They are very polished, very professional, but also kind of soulless. They remind me of illustrations on the cover of a sewing pattern to create a Rockabilly look. Very illustrative. Though the styles the differ slightly, the “look” is the same, and even the couple themselves. Any quirky weirdness has been absorbed into the illustration so it’s barely noticeable. Notice the glasses of the girl in the first pic. They have two different lenses, and it seems parts of the frame have become a ribbon tying up her hair. There is also no overt leopard print. In the first pic, we get cow-like splotches, in the second, something that looks like giraffe. And both pics are orange… very, very orange. It’s a little spooky. It’s also spooky how the engine has concealed the couple’s hands, avoiding the “hand problem” of AI art.
This picture was generated on another engine, and though it’s not as polished, I like it better as art. The couple looks real; they have personalities of a sort. It looks like a quick sketch. The girl even has her hair tied up with a scarf in Rockabilly style.
This pic shows even more personality; I added “Native American” to the mix. Again, it looks like a quick work done by a real artist, but it’s full of personality. I had to fix a few things in a paint program, like the man’s eyebrows, but overall, I like it.
I wouldn’t profit off any of these, or claim them as my own. As far as I’m concerned they are just random gifts from the internet.
I’ll be experimenting more later in these posts, and keeping up with the legal issues.
1 pings