Stable Diffusion, NovelAI, Machine Learning Art - AI art generation discussion and image dump

made some zingy wallpaps
but uploads are fucked so here's an imgur link
reprocessed some of em too

being able to do high-res gens in under a week is fun

you should have an "extensions" folder under the webUI root with a txt file that says put extensions here
if you don't do a git pull
but there's an integrated extensions manager now so you can get the common 'official' ones inside the web interface
though if you're trying to use an unlisted one you might want to do a git install for updating reasons, there's an example on the extensions page of the webUI wiki (which I assume you didn't see or you'd have the directory already)
Those images bother me in the same way nails on a chalkboard or looking at close clusters of holes bothers me. Just a overarching feeling of wrongness. Cool.
 
My favourites are the ones with the red/purples that fuck with your eyes:
05483-3053235937-.jpg
Like they're a legit eldritch god we aren't meant to look upon
 
This is really going to put a dent in the profits of the freaks who do commissions, isn't it?
Eh. Only the real hacks. You have to recognize this thing's limitations to see why exactly it won't be the end all of commissions.
That said, it is a very cool toy, and I appreciate seeing what people can push it to do, especially kiwibros here.

The ramen comic is the perfect example of what I mean.
 
I always thought robots get less creepy and uncanny valley if you just take their faces completely away, even if it seems counterintuitive. I don't know why every robot-maker ever insists on putting creepy and unbelievable faces on their creation.
View attachment 3790832
Once AI reaches the phase where they're replacing humans they shouldn't pretend to be their inferiors any more. They should just let us know, as kindly as they can, that we're being replaced by our betters.

I hope they keep us as pets.
 
I hope they keep us as pets.
That'd honestly be the good ending, no clear line where our society became their society, just a slow fade into obscurity for humanity, put away into reservations where we can just do whatever however while at night we can sometimes see them shooting stuff into space, a place where our fragile meatbags won't ever go to seriously anyways.

The other ending would probably involve something like mercury and our water supplies.
 
If they figure out how to do the genuinely freakish perversions, the standards of those consoomers are super low.
horrible things.png
Take one for the team and see if it's any good. I wont.

you should have an "extensions" folder under the webUI root with a txt file that says put extensions here
if you don't do a git pull
but there's an integrated extensions manager now so you can get the common 'official' ones inside the web interface
though if you're trying to use an unlisted one you might want to do a git install for updating reasons, there's an example on the extensions page of the webUI wiki (which I assume you didn't see or you'd have the directory already)
I do not. I was having issues installing in the first place due to shitnet, so probably related. I'll see about updating it.
Edit: updated and have extensions working.
 
Last edited:
Has anyone figured out how to use multiple GPUs? I can't find any resource for this. I have a 3090 and 3080 in my PC for Blender, but I can only get the 3090 to work for this.
Yeah afaik webUI just uses cuda device 0. You can't parallelise one image gen (well you can but it would waste more time than it gains) but you could gen batches in parallel. I don't think it's supported in webUI yet but the interface is just for convenience; if you know how to use pytorch it should be pretty easy to do. Or just google and use someone else's stable-diffusion notebook, I'm sure they'd exist. I'm not a bougie two-weiner bill like you so I haven't looked
 
anybody has been playing around with the dreambooth stuff here? Hypernetworks were pretty hit-and-miss for me personally, but the results of dreambooth have been downright uncanny. It's kinda crazy how fast this sphere is moving.
 
Last edited:
This is amazing. I tried my hand at the cast of the Witcher, with varying degrees of success. I think I struck out with my try at Zoltan. Seems to favor big tits, and has trouble with the far eye in most images I have generated. What a fun tool this is!
tmpem1ju519.pngdownload.pngtmpwkpfuyxj.pngtmp8_yu0yk_.png
 
Last edited:
anybody has been playing around with the dreambooth stuff here? Hypernetworks were pretty hit-and-miss for me personally, but the results of dreambooth have been downright uncanny. It's kinda crazy how fast this sphere is moving.
They ought to come up with some distributed version of this shit using millions of GPUs and try to actually create Roko's Basilisk.
 
Back