HOW TO RECOGNIZE AN AI-GENERATED IMAGE — Protect human artists at all costs

It’s not news that, despite selling themselves as “ethic” and innovative, AI Developers and software programs such as OPEN AI, Midjourney, Dall-E and more are creating a chaos into the perception of what a real artwork, (those created by people, remember?) or how is valuable and what art done by humans should look like.

Being myself a millennial artist I’ve been trained in a classical way: artistic high school, artists monographies as references, history of art, 3 years of university in children’s book illustration and more, and so I developed a pretty keen eye on how artworks are done.

THAT FREAKING BACKGROUND RING OF LIGHT THAT AI-GENERATED IMAGES HAVE.

And yes, when artistically talented people find their style, they also develop a method: using materials (pencils, inks, watercolour) or advanced graphics software such as Adobe Photoshop and when other wanna-be artists want to master their art they start to OBSERVE (not steal and let AI generate) and even “copying” the ones they admire while finding their way.

SO, HOW TO RECOGNIZE AN AI-GENERATED IMAGE OR ILLUSTRATION:

• those unnatural, exaggerated light
• Disney-looking eyes and faces
• zoom and watch: that goddam sharpness filter
• the “wow!” effect at all costs
• sci-fi super detailed scenarios (where something just doesn’t match)

This suggestion guides are not 100% to be applied literally to every single image, but they are truly recurring elements that, at least, will train your eye not to be fooled by anyone:

Maybe you are an art-enthusiast, or maybe you are a company: will you pay those 100 bucks to the IG guy who calls himself an artist, but who does nothing but writing a line of text online and let it generate images of theft while billionaires fill their pockets?

Or maybe, you are a young artist, and you’re just filled with images from the web which are too beautiful to be true, and you wonder who is behind it, why are they so stunning, and why can’t you be so fast in your progress? Hard to keep up with the 0.5 seconds it takes AI to make 10 images, right?

1. Unnatural, exaggerated light:

AI-GENERATED IMAGE OF A MONKEY

While some human artists may use light as a way to convey tone and style, the AI-generated images are truly generally overusing it, making the “paintings” / “illustrations” look plastic, and, indeed, with artificial look. It’s only a matter of time before they get better, but this seems to be their poor signature.

AI-GENERATED IMAGE OF A CAT WHO LOST HIS LIFE, USED BY AN INFLUENCER TO GIVE AN IMAGE TO THIS GRIEF

Generally, a real artist on commission who abounds so much with lights, would be discarded by an art director or, in general, admonished with the tip of having to rebalance everything.

2. DISNEY LOOKING EYES AND FACES

The Cicciobello plastic and cute face goes on here, bringing a big, well known guy to the game: Walt Disney.

It’s no secret now that he is one of the artists that Midjourney / Open AI used to train their software (see the complete list here, click on “artists” folder) and people love him, right? After all, who doesn’t?

IT SPEAKS FOR ITSELF…

Well, wondering if Disney artists would be happy to see copies of their work and style out there. The trend goes on, and it’s not only to turn your pets into these Disney style images: even “bigger” AI-generators can’t just help it to add some of Disney look here and there.

AI-GENERATED IMAGE OF DEFINITELY NOT ON DRUGS GIRLS

3. that goddam sharpness filter

Just try to zoom a bit, and look carefully at the texture of people’s faces: you’ll notice it, a huge amount of sharpness, alias. the quickest way of a rendering.

DISNEY STYLED SUPER SHARP TAYLOR SWIFT… DONE BY AN ARTIST. JUST KIDDING, IT’S STILL AI.

Artificial Intelligence-generated influencer, Brand ambassador for a Motorsport brand. Probably “hired” thanks to her spotless face.

3) The “wow!” effect at all costs

That is so predictable: like any other thing on the web, AI-images want to surprise you. They want to grab your attention for those 10 seconds before you scroll on. So, human-made art in this case has been dragged and chewed to generate something that’s just “wow!” Next time you encounter one, look closer and start to doubt:

(A hint, AI is a great fan of spaceships)

4) sci-fi super detailed scenarios (where something just doesn’t match)

This last point is strictly related to the one before. Sci-fi elements are part of the wow effect of many AI-Generated images, but sometimes, looking carefully, some details just don’t match. And the artificial collage gets a bit more evident. Human figures are almost absent, but when present, the light and shadow effect as well as the bouncing of light don’t match. It’s like putting a photo of you on your sofa with natural light on a background of a blurred, foggy night with lights here and there. Even when merged, something keeps them separate.

SEE THAT? SEE THE COLLAGE EFFECT? THE BLANKET — THE CYBORG — THE BACKGROUND DON’T FULLY MATCH

If your intentions are creative, there is no need for subterfuge or programs that, as in the case of Midjourney, have taken and used the works of art, if not the entire catalogue of 4000+ artists, HIDING IT, just to train their “machine” to create something identical or brutally manipulated by their unique craft just to entertain people who never had a clue about art itself. The same people that in some cases are even making big money from it.

See the case of the $1500 first prize Art award won by a “painting” made by AI (Midjourney) and presented by Jason M. Allen, and clearly copied from the style of the symbolist Gustave Moreau, which I could recognize since I studied it at school 8 years ago.

Jason M. Allen, “Théâtre D’opéra Spatial” — 1st PLACE AT COLORADO STATE FAIR 2022

Thanks God the image made a scandal, and it was later rejected as a copyright protection.

Poor Moreau.

And speaking of which, I just came across this while searching for the article: Stable Diffusion became a fan of Symbolism as well, apparently.

*Gustave Moreau AI Art Style — Exploring Symbolism — Gustave Moreau Stable Diffusion — Gustave Moreau *(Where “inspiration” and automated image service and “use it for commercial purposes are all in the same page. God, give me the strength.)

STILL UNSURE ON HOW TO RECOGNIZE A REAL ARTWORK FROM AN AI-GENERATED ONE?

• check this tool: metadata viewer EXIF
• Find other suggestions here

DO YOU WANT TO PROTECT YOUR ARTWORK FROM BEING USED AND MANIPULATED BY ARTIFICIAL INTELLIGENCE?

• Check out the webGlaze, the only “glaze” that provides a protective filter on your image, making it undetectable by AI

Here’s my first drawing protected by them, and as you can see, filter is invisible, ✨ not the same for all the flaws of the AI images seen so far. ✨

© Miriam Panieri