AI Art Seething General

  • 🔧 At about Midnight EST I am going to completely fuck up the site trying to fix something.
Honestly, I think artists should be seething even more. So far everyone I've spoken to seems to be very optimistic. At worst, it won't affect them. At best, it will let everyone do "real art" instead of doing boring monotonous tasks, like drawing in-between frames in anime.
It's like they don't realize how big the improvements in this area have been in the last couple of years.
 
Not seethe-related in particular, but I've been running my RX 570 into the dirt on my Linux partition for the past week experimenting. Here's something I managed to get today after a lot of tweaking and multiple passes.
00355-3883851055.0-masterpiece_portrait_of_a_cute_busty_schoolgirl_angelical_face_cute_express...png
The AI software curse, everything is either CUDA-cucked, Winblows tethered or somehow both.
You can do CUDA stuff on AMD on Linux with ROCM. I'd sooner suggest you blow your own head off (in minecraft) than wrestle with all the troubleshooting and hardship that may come with it but it's an option.
 
Yes, it will probably be more difficult for a budding artfag to start making money since all these AI will do it for (basically) free.
It might even kill entire job prospects, making it possible only a small pool of people will actually be able to monetize art.

But on this age of cancel culture, I think artists should be more concerned about people who would have a bone to pick with a specific artist using these tools to make career-ending art and accusing said artist of making it. It doesn't even have to be explicitly wrongthink or nasty fetishes; just make a fan art character in the wrong skin color and watch the madness.
Not that it's impossible to do that before the advent of AI art, however such levels of pettiness weren't exactly worth the time before.

Those with very distinctive art styles should be VERY worried.
 
Not seethe-related in particular, but I've been running my RX 570 into the dirt on my Linux partition for the past week experimenting. Here's something I managed to get today after a lot of tweaking and multiple passes.
View attachment 3683286

You can do CUDA stuff on AMD on Linux with ROCM. I'd sooner suggest you blow your own head off (in minecraft) than wrestle with all the troubleshooting and hardship that may come with it but it's an option.
What is it, Stable Diffusion?

---
The real seething will start when the AI is able to persist new characters. As in: nice purple lady you got here, now we need her laughing, disappointed, eating moon cakes, somersaulting, kicking a tranny in the amhole, etc. Do that, and consumer art is a solved problem.
 
Some of the AI art I've seen has been really beautiful, in a surreal, dream-like way.

I suspect this will end up being like digital art for the people who like to consider themselves artists - a bunch of people with a similar skillset making similar art with the same tools, like everyone drawing the same anime/mamga type characters.

There will certainly be a niche for it as a tool, and some skilled people will make really cool things. I don’t think it will ever really replace traditional art, any more than digital art replaced traditional art. The places where it will be most utilized aren't the ones that would be buying an oil painting.
 
Update on the AI art scene, there are MULTIPLE furry datasets/models derived from e621 and trained with its exensive tagging system via being a booru. The end is nigh. However, it's not trained on underage characters because the programs made to scrape the site cannot scrape images that require you to log in to see them. And the scrapers did everyone a favor and excluded peepeepoopoo tags and otherwise low quality artwork from their datasets.
 
I remember how people claimed how AI cannot produce anything that it's not been trained on, but there's been a development that I learned about quite late, when youtube recommended me this video:
Basically, someone adapted Google's dreambooth model to work with stable diffusion, and now you can train the AI on images of a new object, and order it to draw it in different situations and styles.
teaser_static.jpg

art.png

I was telling people that this would happen, but they didn't listen.
 
I remember how people claimed how AI cannot produce anything that it's not been trained on, but there's been a development that I learned about quite late, when youtube recommended me this video:
Basically, someone adapted Google's dreambooth model to work with stable diffusion, and now you can train the AI on images of a new object, and order it to draw it in different situations and styles.
View attachment 3697743
View attachment 3697749
I was telling people that this would happen, but they didn't listen.
00031-912525837.jpg
(((anti-semitic cat))),[(((screaming))),(((smiling))):0.5], (jubilance), (joy), (anti-semitism), (jews screaming), (rhubarb pie), (((large nose))), Sakai Hōitsu, Carl-Henning Pedersen, Dorina Costras, Cornelia Parker, Arabella Rankin, (((cartoonish))), Celia Fiennes

Negative prompt: ((out of frame)), ((extra fingers)), mutated hands, ((poorly drawn hands)), ((poorly drawn face)), (((mutation))), (((deformed))), ((bad anatomy)), ((bad proportions)), ((extra limbs)), cloned face, ((extra breasts)), (double torso), (double head), ((extra arms)), ((extra hands)), ((mangled fingers)), ((missing breasts)), ((extra legs)),(((signature))), ((long neck)), (((two heads))), (signs), (((text))), (((caption))), ((subtitle))

Steps: 30, Sampler: DDIM, CFG scale: 11.5, Seed: 912525837, Size: 1024x640, Variation seed: 3227657173, Variation seed strength: 0.3, Denoising strength: 0.7

It seems you need at least 12gb of VRAM to start training with your own images on a local SD install.

Too bad you cannot just install more VRAM like you can with DRAM. What a failure of engineering that is. My 6gb of VRAM is irritating enough in its limitation in regards to image generation. To even generate a Full HD image is impossible, even regular HD is pushing it (and often not possible). To think, I never thought I'd use the full 6gb of VRAM for anything ever.
 
Admittedly, some knee-jerk part of me gets mad and anxious when I see people churning out AI art, because I've worked hard to get to my current skill level and still have a loooong way to go, but I'm pretty confident that while it might be good for concepts and brainstorming it can't really manage the fine details. When AI can make a model turnaround sheet, then I'll be scared.
 
Admittedly, some knee-jerk part of me gets mad and anxious when I see people churning out AI art, because I've worked hard to get to my current skill level and still have a loooong way to go, but I'm pretty confident that while it might be good for concepts and brainstorming it can't really manage the fine details. When AI can make a model turnaround sheet, then I'll be scared.
Why would you be scared?
 
AI mainly pisses me off because I was told throughout my life that there are no "shortcuts" to being good at art. You have to put time and effort into it, and that's the beauty of being an artist. Your skills are constantly improving and when you see an artist with really impressive techincal skills. You can admire the years they put into their craft and all the love they put into their pieces. Now some fucking retard uses AI and gets magnificent pieces in seconds? It just isn't fucking fair. But then again, most things in life are like this. You spend years of your life dedicated to what you love, and then some dickface takes shortcuts and is praised and put on the same level as you (maybe even higher)? That's bullshit. AI Art is bullshit.
 
Because I'm still clinging to my foolish dreams of getting into the game art & design industry. I wanna draw and model monsters and landscapes and other cool shit for a living.
Don't worry too much if you have actual talent. There are always going to be people who want art with a human touch. AI however will be a major filter for talentless, low-effort artists, and we may see the end of certain modern styles, like that awful Globohomo flat style (I don't know the real name, but "Globohomo flat style" on your favorite image search engine will turn up more than enough examples). If we have AI that can replicate the style of Vermeer, why would we pay for garbage an AI could have probably done 5 years ago?
AI might make art great again. The human artists who can't do much above flat stick figures will learn to code because an AI will be able to create art that has both the quality and human feeling of their garbage. Truly good artists should have no problem being recognized and well paid because there will probably always be things AI won't be able to reproduce. Emotion, experience, originality, and God forbid the ability to translate such things onto a canvas or image file, are totally foreign to an AI.
 
Surprised there's no AI porn service or porn-specific forks of this.

If I was in porntech (eg: pornhub and such) I would be investing on this.
You'll have to download and run it locally if you want a fresh squeeze of the milk of the jew.
I bet /pol/ would give you a ton of crypto if you keep pumping those pics out.
Update on the AI art scene, there are MULTIPLE furry datasets/models derived from e621 and trained with its exensive tagging system via being a booru. The end is nigh. However, it's not trained on underage characters because the programs made to scrape the site cannot scrape images that require you to log in to see them. And the scrapers did everyone a favor and excluded peepeepoopoo tags and otherwise low quality artwork from their datasets.
RIP bros making bank from furry commissions, those were big bucks IIRC
AI mainly pisses me off because I was told throughout my life that there are no "shortcuts" to being good at art. You have to put time and effort into it, and that's the beauty of being an artist.
Is it? because modern art seems to be getting more mediocre and lazy every year, like they are not even trying to pretend its anything but a tax scam anymore.
 
AI however will be a major filter for talentless, low-effort artists, and we may see the end of certain modern styles, like that awful Globohomo flat style (I don't know the real name, but "Globohomo flat style" on your favorite image search engine will turn up more than enough examples).
LMAO no it won't, the repulsive globohomo art style is pushed by major corporations and NGOs. They don't put that shit everywhere because it's cheap or easy to produce, they put it everywhere because it's the exact shade of soulless they want the entire world to be. The only difference is now the globohomo art will be made by computers.
 
Back