What do you people think about AI art?

[Clone303]

Hmm also heads up places where you can make AI ''art'' for free i know of is NightCafe and OpenArt both places are almost unlimited

[BloodShed]

I think it's fun to faff about with Stable Diffusion.

Though I mostly make weird stuff or use it upscale some old wallpapers (<1024x768) so they don't look grainy/pixelated at 1080p or above.

[Pierce Dalton]

It's an awesome tool for those that can't afford to hire a real artist.

[Haruspex]

As an "artist" who never really got past the middling learning stage, it's a great way for me to take my half-assed art and help it to realize it's actual potential. img2img is amazing.

[The End]

It could be inspiring, but is is a very cold art-style.

Soul-less.


Rex Artis

[Unexpected]

people overhype it as this scary construct that'll put people out of a job, but the reality is AI art is very sub standard and easy to spot as they have very uncanny features about them, i've seen a recreation of duke nukem with 12 fingers, or the gman with 6 fingers on each hand, faces being warped or not proportioned correctly, human art is not going to lose its place

[Dunkler Messias]

It's nice for personal use or if shared with other enthusiasts in a dedicated forum.
But right now people spam AI generated images, which doesn't even look god, everywhere. So it's really annoying for many people. The few ones who likes AI generated images doesn't really care and continue spamming these.

Also bad: There are people who sell AI generated images ... even worse: Some people acutally buy these. It's really bad since most of the images just looks very bad, since many images have obvious marks like 6 fingers on one hand, weird eyes, floating things and many more ... but people doesn't seem to care. :I

[Walach]

Publicado originalmente por Unexpected:

[…]human art is not going to lose its place

Yet. ;P

Anyway, I've been learning to use the stable diffusion ai for over 5 months now and while I can create a lot with it. I believe it's going to be at least a year or more for the ai to get to a master level of creating art. Might take longer for it to "learn" how to create very specifc things. But for the short while it's been here, it's already quite good.

If I had to say something that someone can point and laugh at, then I'd say that the art ai will be "perfect" in about 10 years from now. Perfect in a sense that it can create 98% of all art one can think of, and/or want.

[Mr. Smiles]

It is inevitable, and artists need to learn to live with it and/or use it as a tool, because it's not going away.

[76561198386219967]

future art

[Chompman]

The problem is with it not protected under copyright law then no one will use it commercially and that's a big issue for many.

[BloodShed]

Publicado originalmente por Chompman:

The problem is with it not protected under copyright law then no one will use it commercially and that's a big issue for many.

The only copyright you could really hold is for the prompts and seed.

[AmsterdamHeavy]

The future will be soulless.

[stray shadow]

Publicado originalmente por BloodShed:

Publicado originalmente por Chompman:

The problem is with it not protected under copyright law then no one will use it commercially and that's a big issue for many.

The only copyright you could really hold is for the prompts and seed.

There's no copyright, it's all plagurised from work created by others collaged together, often do poorly that doing a Google Lens search will find the original image it ripped off.

[BloodShed]

Publicado originalmente por stray shadow:

... often do poorly that doing a Google Lens search will find the original image it ripped off.

That would mean the people who prompted those images tried to replicate a specific image, instead of being 'original' with their prompts.

EDIT
OR they're being lazy and doing light img2img WITH an original piece.

[lightwo]

Fun to experiment with. The results can be incredible.

As for its uses, it might fill in the gap when some graphical material is missing and needs to be done quickly. To use it commercially, I'd only ever sneakily generate something here and there not to worry about potential copyright disputes, considering better models are trained on anything, regardless of copyright or license.