pouët.net

AI art in compos

category: general [glöplog]
I think AI art is a nice alternative to stock images. So it might be useful for illustrating articles in a magazine. I wouldn't care, of course, to watch a compo showing the skills of people in finding stock images.
added on the 2022-09-03 13:23:53 by novel novel
Meanwhile: https://pollinations.ai/c/Anything
So, an AI walks into a bar.

The bartender says :"why the long face?"

The AI answers: "How does it make you feel that the long face?"
added on the 2022-09-10 03:15:38 by rloaderro rloaderro
I used Midjourney AI to create some character portraits for an RPG I'm developing in my spare time. The portraits look very good, far better than anything I could draw myself. But I agree that AI-generated images shouldn't be used in graphics competitions at demoscene parties.
added on the 2022-09-10 09:03:04 by Adok Adok
I'm not a graphician but I do tend to judge entries by their creativity/aesthetics and not necessarily by how technical an entry might be or how many hours it took to make it. Of course knowing how something is done is much appreciated.

If AI is allowed in graphics competitions I would hope that people would be specific with the techniques and tools used to create their entries. voters can then try to judge the entries how they see best from their perspective.

Anyways, I asked AI to create competition rules for freestyle graphics compo that allows the use of AI, you're welcome compo orgas:
Code:1. The competition is open to anyone using AI to create freestyle graphics. 2. All entries must be submitted by 11:59 PM on the day of the deadline. 3. All entries must be in JPG or PNG format and must be under 5MB in size. 4. Each entry must be accompanied by a title and a short (100 word maximum) description of the work. 5. A panel of judges will select the winning entry based on creativity, originality, and technical skill.
added on the 2022-09-11 10:37:32 by waffle waffle
where is the "creativity" level then if the "ai" pic is a blended puzzle of scraped pictures from 5 different internet sources? and no its not legal to rip off other peoples work in parts... so the whole "ai art" is not legal at europe.
added on the 2022-09-11 15:33:12 by xeNusion xeNusion
Quote:
Maybe we should have an "I didn't do it" compo?

- Unreal demo engine
- Additional code by Github Copilot
- 2d art by Midjourney and Dall-E
- Shaders copied from Shadertoy
- Music ripped from an unknowing commercial drum and bass artist
- Faked stamps to be returned back
- Cracked by Skid Row
- Minted as NFT


That is what I keep "creatively" joking at and with for years now.. I even occasionally hire top stars (brought back from retirement) to support the model.

I call it demoscenial postmodernism, meaning: nothing new can be invented, so let's have some fun with some already existing crap and satirise it the way, people wonder if this is serious or what the fuck...

I just wished that I'd had either Gargaj's skills who mastered this genuinely with a style or Midjourney's funding so I don't need pretend that much ... ;-)

At the end, we're all swallowing our own tail anyway and for years already.. so what's the buzz?
added on the 2022-09-12 10:34:41 by hollowone hollowone
Quote:
where is the "creativity" level then if the "ai" pic is a blended puzzle of scraped pictures from 5 different internet sources? and no its not legal to rip off other peoples work in parts... so the whole "ai art" is not legal at europe.

Not much creativity involved for sure. But I doubt that it's illegal in Europe. It reminds me of a case where someone wanted to troll GEMA (the infamous German copyright collection society) by registering a 33 seconds track made from 70000 samples from other songs and they basically said "no, you can't register that, only recognizable samples can be registered"
added on the 2022-09-12 12:12:00 by Kabuto Kabuto
Absolutely not trying to play devil's advocate here (since I share most of the opinions here), but here is what a skilled demoscener can do with AI generated stuff:

/imagine - a short slideshow

Cool af imho, even if disturbing a bit.
added on the 2022-09-14 00:25:54 by LiSU^TRS LiSU^TRS
The short version:
The scene should define a competition where AI’s should compete with each other. And another category where developers and the training teams compete with each other. We should not insult talented artists by competing with AI generated images.

Instead the AI realm needs it's own competition. And then AI stuff can truly become exciting in the demoscene!

The long version:
The demoscene has emerged from a history. And I think you can only decide this thing historically. It is called "The Art of Coding". And semantically there are always two levels: The art of getting the impossible out of a big pile of junk by means of ingenious coding. At the same time, however, it is not only about generating "form follows function", but also about creating art from the coding and, above all, from the limitations and shortcomings of the machines. So Art of Coding means to show both the beautiful mind of the coder. And at the same time to show the beautiful mind of the artists. And if we look back at the C64, both the programmer and the visual and audio artists had to work with ugly tools. Like the C64 is very special with its hardware. And the graphic artist always had strange limitations in the pixel editor and the color palette while the musician was struggling with ugly bleeps to turn it into beauty. That's why the contests are always about skills. And even today always exciting.

Today I really struggle with the meaning of certain compos. For example, the PC demo has almost no limits. So what is the competition? Are we looking at high-end films? Just for entertainment? Are we comparing graphic artists? Or do we compare the coding of the shader? If we're comparing the coding, using the Unreal Engine doesn't make sense here. Or is it just about the shaders? But who provided the shader? The devteam of the engine? Or the scener? Sure, the Unreal team gets the credits, for sure. But the unreal team did not upload the demo to the competition server. If this is a competition, what exactly are we comparing?

I think years ago it was clearer what the compo was about. I mean, you can't do pixel art with a 3D editor like Blender. That doesn't make any sense to me at all. From my point of view, a demo team is always a group of skilled artists and toolmakers. Always!

It’s all about Artists and Toolmakers together facing restrictions and searching for the limits of what’s possible. And when a group releases their toolchain, hundreds of almost identical demos appear overnight. As nice as that always looks, what exactly is one supposed to evaluate when voting? For me the teams who created the toolchain will be the heroes forever!

Many demos take a year to finish. And then others contribute stuff done in a day by using a generator. I think for a contest that's kind of ridiculous.

For example, the shader showdown is extremely clear. The tool is always the same. It's also absolutely clear what you're evaluating. Both the mathematical skills and the artistic ones. And this clarity is missing in most of today’s compositions.

It's all about the beautiful mind of the coder and the beautiful minds of the artists. That is the „art of coding“.

And now comes AI. What should we please evaluate there, when we have a voting key? Classically looking at it, we would have to evaluate the coder who wrote the neural network. And the artist who uses the tool. But who is the artist here? And what is the tool? The neural network? Or the already teached network? Is the artists the person who pics the results? Are we now evaluating the "beauty" of the mind of the AI? Then we would actually have to do some kind of shader showdown AI vs Human coder, right? Or Shadershowdown AI vs AI? Because it's still about evaluating the talent. And with AI generated images, I find the talent question rather problematic in gfx compos. Should we now evaluate the talent of the AI? Or what is the point here?

No. For me, the one to have respect for is the programmer of the AI. But if it is an AI from the internet, its programmer is not present at a demo party. Yeah, where's the competition then?

Maybe we should rather separate the categories according to the contest theme.
- Pixelart: Self-made with archaic tools. Very clear.
- Pixelart: Self-Made with a pure pixel editor, free of choise. Also kind of clear.
- Gfx: Choose your tool. But we want to see your skills!
- Tracked-Music: Very „easy“ to compare the skills.
- Streamed Music: Was it a sample pack? Self made? AI-Harmonic Packs? What do we evaluate here? I think this is just about taste today.
- Constrained Palette: Wonderful! Artists really have to suffer with the ugly palettes. Always exciting!
- Sizecode: The Shader is the competition and knowledge about compilers and code. Very clear compo.
- PC demos: Shader-coders and artists compete with each other, right? It’s not about the tools anymore, right?
- AI? Is the AI a new competitor? Or are we letting a digital worker create things on our behalf? Where is the competition? Who is the winner here? The person who did the training? The person which developed the neural network? Or the person which used the AI to let stuff being done? Do we still have competitions at all? Maybe everything has been done on this planet and we don’t need competitions anymore? Computers are too powerful. Every year a new architecture is being released. It seems no contemporary machine is really pushed to the limits anymore, like Amiga or C64.

The art of programming has come a long way. But with AI, it's no longer programming (to me). It’s more a competition in educating. So if Demoscene is doing AI, then AI should be put in its own category where the real competition of the AI realm becomes visible. So the scene should define a competition where AI’s should compete with each other. And developers of AI systems.

But AI generated images should not be brought into gfx compos.

Demoscene is - once again - about comparing talents and not just showing a gallery of beautiful images. Therefore we should not insult talented artists by competing with AI generated images.

Instead the AI realm needs it's own competition. And then AI stuff can truly become exciting in the demoscene!
Super late to the party here... I'd probably ban text to image AI art from freestyle graphics compos, but would allow them in wild, or maybe make a fast compo where everybody has a fixed list of words and to use to pass to the same AI text to image engine and see who comes up with the most crowd-pleasing prompt-to-image.
added on the 2022-09-21 17:22:56 by DrClaw DrClaw
Quote:
I call it demoscenial postmodernism, meaning: nothing new can be invented, so let's have some fun with some already existing crap and satirise it the way, people wonder if this is serious or what the fuck...

I just wished that I'd had either Gargaj's skills who mastered this genuinely with a style or Midjourney's funding so I don't need pretend that much ... ;-)

wot?
added on the 2022-09-24 11:10:59 by Gargaj Gargaj
Isn't that in art compos you have to show steps or you get disqualified? But I guess you could still provide fake steps that are believable and would require less skill to draw than final piece.
added on the 2022-09-26 21:58:31 by tomkh tomkh
DreamFusion: Text-to-3D, is this 4 real?
added on the 2022-09-30 00:30:14 by neoneye neoneye
added on the 2022-10-08 23:47:38 by v3nom v3nom
About as interesting as a 4k upscaler :)
added on the 2022-10-09 17:29:33 by farfar farfar
@nosfe so you asked an AI to generate imaginary prints of lino/woodcuts, then you actually cut the result into a blank sheet of lino and printed it for real ? BB Image
added on the 2022-10-09 19:47:01 by fra fra
Quote:
@nosfe so you asked an AI to generate imaginary prints of lino/woodcuts, then you actually cut the result into a blank sheet of lino and printed it for real ?


yeah like that, going to do a series of prints based on that idea.
added on the 2022-10-09 20:00:29 by nosfe nosfe
Quote:
yeah like that, going to do a series of prints based on that idea.

awesome!
added on the 2022-10-09 20:27:02 by fra fra
Outside of the ethics, i quite love many AI art pieces which blend different artstyles together in a way a human wouldn't.
IDK. In my mind this type of usage is respectful to the original artists.

About AI-only compo - I think it's a bit lame as a concept. But I like the idea of using it in your demos tastefully.
added on the 2022-10-09 20:57:00 by wrighter wrighter
Working in the demoscene mostly as an artist - of course I tend to see this pure usage of dall-e, midjourne and so on but with a technical interest also curious.
But yeah - that pandora box is open ,now, No way to bring that ghost back.

But the fascinating stuff for me are here - and just can scratch them here in a forum post just on the surface.

- technical aspect
- psychological
- inspirational
- some tool aspects

to define that more (for me and myself) in detail:

- technical aspect:
The same part of the fascination of demoscene - what is possible - same here. Reducing it to gfx is just modern - but not true.
it was just first one which makes it visibile and with that fascinating to people without a technical or creative artistic background.
But as anyone can set something up - it would be interesting in a compo what else could be possible if people tweak it in a technological way.

- psychological:
the usage explains a lot about us. like the used prompts, like the results from the ai itself and because it has so far trained on human made stuff and show thatin its results.

- inspirational:
if you see the psychological aspect- it can help to adjust your own atitude to some topics for your own ideas if we leave it gfx aspect.

- tool aspects:
if we leave the features to helper functions. These could speed up alone an creative process enormously just to facilitate time consuming aspects.
But well, this might be a matter of a other discussion.

For more I wouldn´t use it - as I can use graphic tools and paint even in my humble "personal" style. Well, I already learned them.
Even I sometimes use "inspiration boards" if I am working on some special topics with a theme, but the creative process itself is what is what makes my personal fun and freedom
and where I get my satisfaction from. I still hope for future generations it might be the same - but as the usual behavior of humans as a mass tend to be that of an dummy user (fast use and consume adn don´t think how)- we will see what it brings.

But back to orginal question - ai art in compos:

I was also thinking first it would be good idea to split the gfx compo into to parts to keep the ai art seperated.
But that is in regards to the 4 aspects I pointed out above, now I think it is not a good idea. Because it is thought too small.

- Like I wrote AI usage is already in more places than just graphics - it is just the most obviously one.
- Fakers Paradise. Retouching a Ai image with photoshop and then faking steps ? wouldn´t be prob. We have already to many good copycats and converter boys out there.
- it is copied from existing art (just let alone the prompts with artist names it is just a matter of time for upcoming changes in copyright, personal rights and creative licenses -
that will be an interesting topic for itself for the future)
- to the current time you can identify most ai generated images if they are unmodified from some artists hand.
That is already changing these whole public beta testing of midjourney and Dall-e and so, trains the ai behind it, too.
It is amazing how many people each day are using it online. it is evoloving with that data, too. let us hope the most users are tending to use same cliches - I believe I can see some repeating color palettes schemas in these image creations already. ;)

My personal conclusions:

If there should be an competition for ai - don´t reduce it to gfx - change it to a technological one. would be interesting and could bring people/public interest into our scene.
For gfx competition itself is just a matter of time you won´t see it anymore in the final result.
So the rule "not to use ai for creative content" and steps should be enough- yeah there will be fakers. There have been copycats, convert boys and the ai smoothers will join them.
Next target for ai content will be music and if not at the same time, after that we will have ai generated demos out there.

On the other hand: it is also a call to do more live (drawing? music? ...) compos in this combination. If I see that someone can do something in realtime, I also can believe he can do it in a competition.
Not so good for some of us old lazy bags that just wanna drink and hang around on parties ;). But would also nice to see.

So keep an open mind and be also a bit critical - I just want to add at the end some words to this topic from Marshall McLuhan:

“When we invent a new technology, we become cannibals. We eat ourselves alive since these technologies are merely extensions of ourselves. The new environment shaped by electric technology is a cannibalistic one that eats people. To survive one must study the habits of cannibals.”
added on the 2022-10-10 11:44:32 by titus^rab titus^rab
I'm going to stir this up again due to a recent fuzz at Artstation where people are mad that AI content rises to the top of the explore list and Artstation's poor handling of the "crisis". Why people want to submit AI stuff is a matter of guessing of course but sociopathy or financial gain (the hope that prints will be ordered) might be the reason for some.

It has been talk in this thread of AI art being "better" (technically) and "never before seen". This might be but art is not just about technical skill, it's fundamentally about communication. What does the AI and the Artstation AI submitters want to communicate? Not much I dare say.

I saw a funny YT video where the AI generated the same blank anime faces no matter what crap you threw at it. I understand not all AIs work like that but the tech is full of problems that's for sure.
added on the 2022-12-15 19:19:05 by El Topo El Topo
Quote:
This might be but art is not just about technical skill, it's fundamentally about communication. What does the AI and the Artstation AI submitters want to communicate? Not much I dare say.


the same thing a regular drawn image communicates.
what is your point exactly?
an image communicates what the image communicates. you think it has no reason for existing and sharing? that it's random word selection and random decision to upload it? even if that was the case (which it's not) it would still be communicating something.
so what's your point?
added on the 2022-12-16 11:32:20 by psenough psenough

login