That was my intention. I used the studio ghibli, 1980s (style) and cel shading tags (among others), along with a low seed, to get that style.
Wait... are you saying lower seeds correlate with older images?
Well I noticed all that. I just was confused if the low seed was similar to those. I don't really understand "what" exactly it refers to, I just know it indicates some base structure or something.
An AI generating images like this feels so wrong
If you think that's
surprising, you should check out some of the crazy shit
people have done with Midjourney.
Maybe? I'm not sure how to replicate that particular style, and it is certainly a very good example, but otherwise it doesn't look too
far outside of what you can generate with the standard models. Here's my (fairly half-assed) attempt to get something similar out of NovelAI for comparison.
what the hell
I can easily find myself following an artist that draws this......
Stable Diffusion's weights have it too
FUCK that is good and I hate it
in a sense, it being one of a kind, the artstyle and everything, is novel; on the other hand I probably won't ever see another art with this artstyle and I feel that regretful
shewont be chibi for much longer
Nice! More please!
wonder why stable diffusion images all have this watercolor paper texture
Well, a chunk of the training data was on traditional art and apparently one of the go-to phrases that people use to improve the basic models is to add "masterpiece" which seems to invoke them.Fran
in Yamato! I know what you're thinking, but I can't seem to get my Aquaplus thing to NOT produce breasts. Let's just say it's the magic of Utawarerumono. Also after like 100 generations this was the least suggestive pose. I give up.
Argh I just now noticed the arms and hands.
It's weird, I'm so focused on the faces that everything else is just background noise, including what appears to be literal background noise on the left
I'm impressed by just how well the AI was able to draw Fran on the right. Was it somehow trained on some of her art?
Yeah, I downloaded an embedding of her that someone made from training on screenshots of her from the anime.
This prompt I threw together somehow keeps attempting to make video game screenshots. Cool.