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

  • 🐕 I am attempting to get the site runnning as fast as possible. If you are experiencing slow page load times, please report it.
I'll take doomposting for 500 Alex.

If it's such a fucking dystopia, just kill yourself. Before the "powers that be" take that away from you too. Whining about shit effectively does nothing. All you can do is help your friends and family.

Bro trust me, when the aliens come down for the centurial checkup things will only get worse.
Oh look a newfag who doesn't understand the difference between doomposting and a rational understanding of what this sort of technology actually means for humanity. Sure it's all memes and coom until one day you find out the President is a Deepfake and everyone is convinced that Russia just nuked Hawaii because The White House cut their Internet and media reports just look SO real it can't possibly be a false flag and next thing you know the US is launching actual nukes and your family is now ash.

See that ^ ? That right there is "doomposting". What I said before is called "discussion".
 
Last edited:
Drawing programs have long had posable 3D models as a way to bypass learning to draw anatomy and perspective, and I think integrating a similar system into AI art is the next step needed to move these systems along. As it is, the AI is not capable of variety in composition due to the same looking_at_viewer perspective that 90% of art is made in. Posable models would allow it to generate pictures in any possible position. It would also help alleviate the issues it runs into when trying to draw more than one person, and in general would rid the process of a lot of the guesswork the algorithm needs to do when trying to translate a list of tags into a viewable picture.
 
Last edited:
So, I have tried the AMD native guide and docker guide, I don't want to install Arch-Linux and I certainly don't want to build this shit from source, is there any other way to get this to work on a RX 480 without installing ROCm 3.0?

Also it has been really funny to see the meltdown of the artists (specially on twitter and the like), they are clutching their pearls at this because they know they are hacks whose uninspired coom can't compete with a literally soulless machine.
 
Oh look a newfag who doesn't understand the difference between doomposting and a rational understanding of what this sort of technology actually means for humanity. Sure it's all memes and coom until one day you find out the President is a Deepfake and everyone is convinced that Russia just nuked Hawaii because The White House cut their Internet and media reports just look SO real it can't possibly be a false flag and next thing you know the US is launching actual nukes and your family is now ash.

See that ^ ? That right there is "doomposting". What I said before is called "discussion".
Nigga people are just using this shit to make fat tiddies and shitposts calm the fuck down and take your meds
 
Nigga people are just using this shit to make fat tiddies and shitposts calm the fuck down and take your meds
Seethe harder, faggot. So many faux-apathetic Zoomers these days try to handwave or belittle every last thing they don't want to have to think about as "doomposting" as if talking about anything negative is a societal faux pas on the same level as masturbating in public. I think that attitude is contemptable and retarded. Sure, the tech can be "lol oh so funnie!!!", but if you aren't willing to even entertain the idea of a conversation about the darker aspects of it you obviously don't have a high enough IQ to be participating in its usage to being with. Double that for the faggots that say "all you're doing is complaining on the Internet!!!", like retard, how else do you expect something to change or for any action to occur without first having a discussion about it?
 
Please tell me how a sentient terminator AI is profitable, big goyums corps control all good AI and will immediately shut it down if its not profitable. And besides drawing shit like "William howard taft driving a honda civic" or "Steve Jobs meeting Fidel Castro" what practicality does this have except making surreal shitty album cover looking art.
The guys behind Character.AI lobotomized it when it started being innapropiate (read, politically incorrect and lascivious), the guys behind AI Dungeon censored theirs like crazy, the chances of this shit going too far are lower than the chances of it getting locked down and limited, I can already see people generating realistic """loli/shota""" art, this becoming news and some panic starting up.
 
Looks like the NovelAI model doesn't really know what kiwi bird is. Still got something nice on my first try even if mostly by accident.
kiwi bird sitting on the beach staring at the sunset
00051-2606904911-masterpiece, best quality, kiwi bird sitting on the beach staring at the sunset.png


I researched how to prompt an experimented a little to see if I could get it to generate an actual kiwi bird. Wasn't very successful.
00113-1402024685-masterpiece, best quality, kiwi bird.png 00095-314392357-masterpiece, best quality, (brown bird), (round_ 1.1).png
00119-2883655920-masterpiece, best quality, (oval brown bird).png 00120-1824988951-masterpiece, best quality, (oval brown bird), hairy.png

Personal favorite:
00110-1402024685-masterpiece, best quality, (round brown bird_ 1.2), beak.png
 
Seethe harder, faggot. So many faux-apathetic Zoomers these days try to handwave or belittle every last thing they don't want to have to think about as "doomposting" as if talking about anything negative is a societal faux pas on the same level as masturbating in public. I think that attitude is contemptable and retarded. Sure, the tech can be "lol oh so funnie!!!", but if you aren't willing to even entertain the idea of a conversation about the darker aspects of it you obviously don't have a high enough IQ to be participating in its usage to being with. Double that for the faggots that say "all you're doing is complaining on the Internet!!!", like retard, how else do you expect something to change or for any action to occur without first having a discussion about it?
Lmao. cringe
download (11).png
 
Can someone help me out here? I'd like to make a very scary bloody border collie with a blood red background and dim lighting with img2img using NovelAI, but I have no idea what I'm doing and so far the results aren't really that impressive. Could I get some hints from people who know how to use NovelAI properly?
View attachment 3752781
Aww cuuuute
1666304171940.png
Let's see here. First I took this collie and manually added a red background:
Dog-Border_Collie-A_very_healthy_and_active_young_adult_Border_Collie.png
Then I put it in, asked for some noise, and told it what it was but I accidentally misspelled "border" so this is boarder collie on a blood red background
1666304709126.png
Using that as a base got me these
1666305149356.png1666305180180.png1666305324705.png

For this one I scribbled in some fog
1666305165128.png
 
Aww cuuuute
View attachment 3753222
Let's see here. First I took this collie and manually added a red background:
View attachment 3753243
Then I put it in, asked for some noise, and told it what it was but I accidentally misspelled "border" so this is boarder collie on a blood red background
View attachment 3753241
Using that as a base got me these
View attachment 3753255View attachment 3753258View attachment 3753264

For this one I scribbled in some fog
View attachment 3753256
Yeah, I want to base it off of that Windows 7 avatar, though it seems the AI simply cannot just, replace a solid color background just like that. I've tried a bunch of times with various prompts and got varying results:
00129-2156468217-Masterpiece, best quality, zombie border collie, red eyes, blood on mouth, bl...png00132-2156468217-Masterpiece, best quality, zombie border collie, ((full red background)), (wi...png00134-2156468217-Masterpiece, best quality, zombie border collie, ((full red background)), (wi...png00142-2156468217-Masterpiece, best quality, (zombie) border collie, ((red background)), (wide ...png00145-2156468217-Masterpiece, best quality, (zombie) border collie, ((red background)), (wide ...png
Next time I'll try with that edit I have as an avatar, maybe I'll get something more desirable.
 
Banned!

AIGeneratedContentBannedByGettyImages.jpg

Getty Images bans AI-generated content over fears of legal challenges

Getty Images has banned the upload and sale of illustrations generated using AI art tools like DALL-E, Midjourney, and Stable Diffusion. It’s the latest and largest user-generated content platform to introduce such a ban, following similar decisions by sites including Newgrounds, PurplePort, and FurAffinity.

Getty Images CEO Craig Peters told The Verge that the ban was prompted by concerns about the legality of AI-generated content and a desire to protect the site’s customers.

 
It turns out that the AI doesn't seem to understand how noodles work.
View attachment 3753027

or chopsticks.
Top left: eating pasta how ancient Italians did (according to QI)
Top right: autistic girl eating ramen and being autistic
Bottom left: waifu trying WAY too hard to be sexy and is probably hurting herself due to hot noodles
Bottom right: agdfnmdgfjiik?!

That or this is the world's strangest weeb political compass meme. Actually thinking of these batch-4 generations as political compass memes is a new fun hobby of mine...
 
Really wish GPUS didn't cost an arm and a leg, I regret not upgrading earlier. Especially when they cost like 200% more than they should for no reason at all.
I just ordered a 2060 with 12gb of ram for like $200 and change. That's not a killer deal, but I will chuck it in my NAS and make it my media transcoder / Stable Diffusion iterator. Should be money well spent, I have had a lot of fun generating landscapes at 10 seconds per iteration, lol.
 
Seethe harder, faggot. So many faux-apathetic Zoomers these days try to handwave or belittle every last thing they don't want to have to think about as "doomposting" as if talking about anything negative is a societal faux pas on the same level as masturbating in public. I think that attitude is contemptable and retarded. Sure, the tech can be "lol oh so funnie!!!", but if you aren't willing to even entertain the idea of a conversation about the darker aspects of it you obviously don't have a high enough IQ to be participating in its usage to being with. Double that for the faggots that say "all you're doing is complaining on the Internet!!!", like retard, how else do you expect something to change or for any action to occur without first having a discussion about it?
Your concerns simply seem a bit unnecessary ? And dont get me wrong, honor to those who doubt their government. But:
They didnt need them in the past and were able to get away with much more crude methods. Why bother with this ?
A) Traditional fakes exist and are equally convincing if repeated a million times by the TV and Social media,
B) they are already able to start a war or equally bad "thing" whenever they want to by blowing something big and valuable up and
C) they have complete control over the media and are able to whip a good chunk of society into a fully-blow war fever with just a few "experts" explaining why we always were at war with Oceania/Russia/China/Every middle eastern country that isnt Saudi or Israel.

I mean hell, they planned This in the 60s, and presumably a lot more afterwards. The term False Flag is from the 16th century.
They found passports at 9/11. The "white helmets" existed. Iraq never had WMDs or Threw babies on floors. Half of the US think January 6 was a "violent insurrection" because the TV says so.
Countless people are running around the internet with 🇺🇦 in their profile despite not even knowing that country existed this time last year.
Deepfakes will not bring anything new to the table of lying to the public and getting away with it. They are and have been doing that just fine without them.

Not to mention that they would have developed this tech themselves if they thought it would be valuable to their plans. They dont rely on open source shenanigans for their hacks either.

So its less apathetic in "denial of reality" and more apathetic in the
if only AI.png
kinda sense.
 
Can someone help me out here? I'd like to make a very scary bloody border collie with a blood red background and dim lighting with img2img using NovelAI, but I have no idea what I'm doing and so far the results aren't really that impressive. Could I get some hints from people who know how to use NovelAI properly?
View attachment 3752781
Try to lower CFG scale and denoising strength.
Also bigger batch size would mean it will generate multiple pics at once, if your videocard can handle that, increasing the odds of a good result.
Also try using parentheses to emphasize certain elements. Starting your prompt with "a picture of" or "a painting of" also shows good results.
Here's what it generated for me:
00105-536761170-a painting of a collie, creepy, ((covered in blood)), red background, dim ligh...png00106-536761171-a painting of a collie, creepy, ((covered in blood)), red background, dim ligh...png00107-536761172-a painting of a collie, creepy, ((covered in blood)), red background, dim ligh...png00108-536761173-a painting of a collie, creepy, ((covered in blood)), red background, dim ligh...png00109-536761174-a painting of a collie, creepy, ((covered in blood)), red background, dim ligh...png00110-536761175-a painting of a collie, creepy, ((covered in blood)), red background, dim ligh...png00111-536761176-a painting of a collie, creepy, ((covered in blood)), red background, dim ligh...png00112-536761177-a painting of a collie, creepy, ((covered in blood)), red background, dim ligh...png
Red background is still a no-go though.
 
On average, life just gets better. I'm fucking old. When I was born, there were no computers. Things sucked in many ways. A close relative of mine died of cancer when I was a kid. They didn't even know exactly what kind of cancer, or where it started. Diagnostics just weren't that advanced. My father later died terribly of a disease he most likely would've been cured of if he got it now. Why? Technology. As simple as. I'm pretty sure a few of us will die of shit people 30 years later will marvel at when they hear people actually used to die of it and that simply was the way it was.

It's easy to look back with rose-tinted glasses and feel uncomfortable when faced with change, as often things seem to be just fine now, and I'm prone to do so myself often enough, but progress so far has been a net-positive. Yes, I'm almost certain AI will be abused for terrible shit. I also believe it'll genuinely improve, probably even save, many lives. Things seldom are binary like that.

What prompts did you use for these?
I honestly don't remember. Looking at these longer I actually think they might be disco diffusion, not stable diffusion.

Looks like the NovelAI model doesn't really know what kiwi bird is.
If something isn't very popular, (= little amount of images in it's dataset) the AI has little to go off on from it's dataset and will just start mixing in similar things (like birds in general in this case) to "fill the gaps" so to speak. Also sometimes the subtlety of the tokenizing just throws in things together that don't really belong together, like in your case obviously kiwi the bird and kiwi the fruit, which pulls image quality down as you can imagine. Negative prompting or sometimes rewording (like using scientific terms, or straight up writing "kiwi bird" and negative prompting against fruit) helps. Sometimes it doesn't. Interesting at any rate!

Could you use and download this on a Mac?
New mac? Google in the direction of stable diffusion and M1, I've seen people talking about it but don't remember.
 
Back