Maybe it’s just me but I don’t find “Loab” all that interesting. So you found a local minimum in the latent space? And when you mix the local minimum with other points in the latent space it’s still kind of looks like the local minimum? Big deal, of course it does.
I guess my other point is, of course there’s a creepy death mask woman in the model. It’s a model of human art. And yes it is creepy and spooky. We are humans.
And it is even weird that it shows up at a negative prompt, I guess! But what is the big deal about it “reappearing”?
Maybe I need to try it myself. It sounds like the artist was remixing the original Loab image with other images and finding her in the image. And that she was still appearing in images several iterations downstream.
But isn’t this just what would happen with any character?
Basically I don’t think this is a feature of math, I think it’s a feature of the dataset.
And of deliberate remixing, which is art and creepypasta but not space demons.
🙂 yea
No it’s just some shit they called it. Y’all are getting slendermanned
lmao that’s a great analogy ty
I don't understand a single thing @deepfates and @mattskala are saying about algorithms but I was scratching my head over that Loab thread bc if you keep adding pepper to your food it's going to keep tasting peppery??? I didn't get what was surprising. Still some fun art tho.
— Quinn 🏳️🌈 (@ContrarianQuinn) September 8, 2022