How do you tell when something is AI art?

[UngainlyTitan]

Wild. Not with the ones I’ve seen. It’s a fresh start each time. You have to refine the prompt and roll the dice. You can’t take an already generated image and iterate on it like that. What program are you using?

Ok, I cannot definitely say that it does not redo the picture, but I did play with the bing AI and explicitly instructed to reuse the image but change thus. It appeared to work, but I could not swear it was working as I thought.

 

[Thourne]

I haven't messed with it much in several months but, stable diffusion would let you take an image you'd already generate and then give prompts to alter the image. I don't remember it being able to do a fantastic job but I did use it to change hair and such.

 

[overgeeked]

Ok, I cannot definitely say that it does not redo the picture, but I did play with the bing AI and explicitly instructed to reuse the image but change thus. It appeared to work, but I could not swear it was working as I thought.

What prompt would you use? “Take that picture and make their eyes wider” or something?

 

[schneeland]

Can you tell the AI something like "I want a fairy sitting on a toadstool?" Then once the picture is done, fine tune with prompts like, "Widen the eyes a bit" and "I want the wings to be shaped like hawk wings instead of butterfly wings" and so on until the picture has met your mental image?

There is a feature called "inpainting" that you can use to refine existing images with further prompts. I heard from some people that they had success correcting areas of an image that turned out to be not to their liking, but I haven't used it personally yet.

 

[UngainlyTitan]

What prompt would you use? “Take that picture and make their eyes wider” or something?

That is what I did. It appeared to work for a couple of integrations. After about 4 iterations it started to drift a lot from the original image. Also I found that the AI did not understand instructions about interactions with objects in the image.
I wanted an image of a sowrdmage elf stepping out of a portal.
It took a couple of goes to get the right elf image but I could get the image before or behind the portal it could not manage the concept of stepping through.
That said, I did not try too hard but I was impressed at what it managed.

 

[Jfdlsjfd]

Can you tell the AI something like "I want a fairy sitting on a toadstool?" Then once the picture is done, fine tune with prompts like, "Widen the eyes a bit" and "I want the wings to be shaped like hawk wings instead of butterfly wings" and so on until the picture has met your mental image?

That's possible. There was an early image-to-image model called instruct-pix2pix that worked this way, but it wasn't updated and is now way behind the times. There are other ways of having the generated image matching your mental image that are (currently) easier than prompting modification in natural languaege, of example inpainting (where you select a part of the image with a brush, in your example the wings, and tell the model to draw hawk wings on it, and it try to mix the new prompt with the existing image, with the hawk wings where you want them), or tools like gligen where you draw boxes on a canvas and tell the model to generate specific things in each box, so you can get better composition of the image and not just random elements all over the picture.

If you can, then art created like that would for me cross over from AI created to artist created via an AI tool. The ultimate piece of art is the artists vision, not whatever the AI decided to draw upon for the original picture.

True.

But the question was "how can one tell if an image is AI-generated or not?" and it's often more apparent at very high resolution. I am pretty sure in a printed book with few details, it is already impossible to tell an AI work (well made, touched up for realism, not just a "type a prompt and click generate until you're cool with the result", which is quicker but will keep artifacts) and a graphic-software using human. It is very hard for photorealistic picture already, and the telling signs are tiny (slightly deformed reflection in the irises, for example, or errors in the fine detail of a cloth pattern... but it only works to detect AI-generated images passing as photographs, as illustrators wouldn't generally include such a fine level of details in their drawing anyway) . If you stay in the realm of fantasy pictures, how can one say that the pattern on your butterfly fairy wings are not realistic enough? There is more leeway for creativity, and most of the signs at this point can pass as artistic choices.

Arguably, consistency is difficult right now (but it won't stay true in the next few month) so asking for one character in a series of scenes could help improve detection. But if it was a "contest" and the submitter knew he'd have to pass the scrutiny, he'd probably train the AI on his first creation and tell it to reuse it for the next image. It is becoming very easy to transfer a specific face to improve consistency.

I don't think at this point it's possible to have an overall "test of AI-ness" working everytime. It would depends on the specifics of the art piece (mostly style, adherence to composition and very little details), and thus "AI-art" would need to be narrowed for a satisfying answer to be determined. After all, if one draw a t-rex and tell the AI to put it in different backgrounds, it would be AI-art but identifying it would rely on focusing on the background, not the T-Rex. But I guess you'd say it's not AI art, so it should pass the test of being man-made.

The areas where the generative AI is making mistakes are narrowing and while it was easier to detect in the long forgotten past of 2023 as "bad hands" is something that can be apparent on a most picture, the current difficulties are much less general. Specific interactions between items are hard, especially if the interactions are uncommon and the AI was rarely trained on image representing it. It can have someone eating with chopsticks, for example, but if you ask for someone being in the process of eating a burrito, there is a good chance the AI will draw someone cutting it and not eating it the right way. Same with sushi: it will most probably draw someone eating it with the rice at the bottom instead of turning it so the fish slice is facing the tongue, as one should. So one could commission a specific image of someone eating sushi correctly and a real human illustrator will be able to do it, while AI will most likely struggle with it. But it's impractical as a test, because one can't draw a whole monster manual's worth of monsters eating sushi the right way.

 

[overgeeked]

That is what I did. It appeared to work for a couple of integrations. After about 4 iterations it started to drift a lot from the original image. Also I found that the AI did not understand instructions about interactions with objects in the image.
I wanted an image of a sowrdmage elf stepping out of a portal.
It took a couple of goes to get the right elf image but I could get the image before or behind the portal it could not manage the concept of stepping through.
That said, I did not try too hard but I was impressed at what it managed.

Gotcha. For something like that you’re probably better off building it piece-by-piece and assembling it in layers in Photoshop or GIMP to make it look right.

I get better results giving a vague prompt and being surprised. The chances of it showing me exactly what I want are infinitesimally low. The chances of it surprising me with something kinda cool are much, much higher.

 

[Distracted DM]

I look through art almost every day for my games, and I detest the dross I have to sift through since AI art got good enough to look pretty good at first glance.

At the moment I can usually tell by the style, or by looking closely, but that's hardly exact. There's definitely a few pieces that I thought were AI until I reverse-image searched it and found it was posted in 2017, or that it's on an artist's social and they don't do AI etc.

The best way? Do a reverse-image search. You won't get a lot of hits, the ones you do will be on an AI art-gen site, and none of them will be on an artist's socials etc.
When that doesn't work anymore, I got nothin'.

 

[Zardnaar]

Can't reliably tell unless it's kinda obvious. Eg fingers, fine detail in the the back ground, some hands and feet.

Quality wise some of it is better than human.

So unless it's hand drawn or painted......

Uncanny valley effect is on a lot of digital art as well.

 

[Andvari]

I would say that becomes a thing when the AI starts spontaneously producing art. As long as it is a mechanical response to a prompt or set of prompts the art originates with the human.

This is trivial for them, but it’s a waste of processing power to have an AI produce arbitrary art no one is asking for.