OT: AI Graphics: Hey @DirtyRU, you using AI yet?

[GoodOl'Rutgers]

For anyone interested in AI art and they have a gaming capable PC (big honking video card).. I have been playing with Stable Diffusion (which is an AI that runs on your local PC, though you can run it in shared servers like Google Colab) for some time now and have recent need to generate a paterned background and have been astounded by its abilities. I know there a lot of virtual GF and NSFW stuff and especially Anime stuff going on out there.. maybe because there's few billion more chinese men than women... but I have bene toying with "photoshop" type editing using it.

I called on @DirtyRU for this because I have noticed his excelelent use of varisu fills and effects in his Rutgers art here. some quick examples.. each took about 20 seconds to generate once I got the prompt doing more what I wanted..

(EDIT.. links from imgbb did not work embedded.. in reply below I placed the direct links.. but here is a IMGGUR post of several of them... all generated from textual descriptions of various complexities.. while you can download backdrops to use with your cutouts.. if you have teh gear to do video editing and play high end games.. you can generate your own art to use in various projects.. even combine it with your real-world imagery. There are "free" online sources where you use their servers.. but, more and more, you need to subscribe to them to get anything done.. and, certainly, on a free plan I would never have gotten to this point with this prompt while generating only 10-20 images a day for free.)

 

[Rufaninga]

For anyone interested in AI art and they have a gaming capable PC (big honking video card).. I have been playing with Stable Diffusion (which is an AI that runs on your local PC, though you can run it in shared servers like Google Colab) for some time now and have recent need to generate a paterned background and have been astounded by its abilities. I know there a lot of virtual GF and NSFW stuff and especially Anime stuff going on out there.. maybe because there's few billion more chinese men than women... but I have bene toying with "photoshop" type editing using it.

I called on @DirtyRU for this because I have noticed his excelelent use of varisu fills and effects in his Rutgers art here. some quick examples.. each took about 20 seconds to generate once I got the prompt doing more what I wanted..

generated at 512x512 just a pattern radial gradient sunburst

generated at 1024x1024 sloppy impasto oils

generated at 1024x1024 palette knife look

even the mistakes along the way were interesting more often than not..

Your images are not rendering for me to view

 

[GoodOl'Rutgers]

Your images are not rendering for me to view

aww.. imgbb.. Ill look again.. these are the links... guess I need to try IMGGUR or something.. I'll edit post above as well.

https://i.ibb.co/gPL2wzp/Comfy-UI-00299.png

https://i.ibb.co/vZ58Nk8/Comfy-UI-00384.png

https://i.ibb.co/SfBPf4Y/Comfy-UI-00483.png

https://i.ibb.co/gtDGd6n/Comfy-UI-00058.png

https://i.ibb.co/bH2yMx0/Comfy-UI-00356.png

 

[DirtyRU]

Cool! I dig' em. Re: me, for the most part I've resisted the urge, in large part because AI infiltrating my industry is the first step towards a guy like me becoming extinct, ha. That said, the graphic design industry has been dominated by Adobe apps for decades, and they have worked some AI functionality into their software, which I've dabbled with just slightly. Obviously I respect and admire AI's capabilities, and have seen some terrific artwork as a result, but yeah, I'm slow to plunge headfirst into it. Eventually I will. But right now, it's also not really "necessary" for the artwork I execute, both professionally and freelance/spare time. At this point it would be just for fun to explore its capabilities.

 

[RuSnp]

Cool! I dig' em. Re: me, for the most part I've resisted the urge, in large part because AI infiltrating my industry is the first step towards a guy like me becoming extinct, ha. That said, the graphic design industry has been dominated by Adobe apps for decades, and they have worked some AI functionality into their software, which I've dabbled with just slightly. Obviously I respect and admire AI's capabilities, and have seen some terrific artwork as a result, but yeah, I'm slow to plunge headfirst into it. Eventually I will. But right now, it's also not really "necessary" for the artwork I execute, both professionally and freelance/spare time. At this point it would be just for fun to explore its capabilities.

The text to image models are getting better and better but I still wouldn't compare them to the stuff Dirty does. They are more idea generation/when not quite professional is fine (like a blog post etc). But if you want a professional grade thing it still pays to use an expert.

 

[Rufaninga]

I found generating a scene or image was pretty good & well beyond my graphic abilities. But try to get something specific in that scene was much harder to have it generate what I wanted. I was trying for things that were more realistic vs illustration based. Also my images were based on my prompts, which garbage in garbage out.

 

[GoodOl'Rutgers]

Yeah.. I know.. Still no substitute for skill and a good eye and creative imagination. BUT.. The specific reason I mentioned this re:dirtyru is because everyone needs fills and masks and stuff.. And often they come from manipulated photos or geometric patterns.. Time consuming stuff or you buy em from a stock source or a free library.

I was playing with AI for the nerd factor and a couple friends were doing a craft project and I suggested I could print a pattern on some paper as a backdrop in the frame for their art. So I searched web.. to borrow and manipulate and I thought.. Hmm AI for just the backdrop? No main object? No naked celebrity? No politician looking ridiculous?.. It worked too well.. Sort if.

But final solution was just the polar vortex trick to make a sunburst pattern in vectors then change color fills. Dudnt use the AI output. But it was interesting.

 

[GoodOl'Rutgers]

I found generating a scene or image was pretty good & well beyond my graphic abilities. But try to get something specific in that scene was much harder to have it generate what I wanted. I was trying for things that were more realistic vs illustration based. Also my images were based on my prompts, which garbage in garbage out.

The new ComfyUI model called FLUX is very very good. Consistent characters have been beyond me for awhile but I got it first try with flux. I think company is black forest labs or similar name.. It's so good that company has to get raided for talent or get bought out.