Can Artists Spot The AI Art?

In this video I challenge artists to see if they can tell the difference between ai art and real art.
My Patreon: / skynixart
discord server: / discord
instagram: / skynixart
Artists in the video:
Monitor Comics: / @monitorcomics
lightofthedeep: / @lightofthedeep
wispravine: / @wispravine
Sebs or somethin: / @sebsorsomethin
Artists that had artwork featured in this video:
yukuri130: / yukuri130
Svetlana Bukanova: www.artstation.com/sayomi
Keyyuui: keyyuui?lang=en
omj_tenco_bot: / omj_tenco_bot
somenormalartist: somenormala...
GUWEIZ: / ttguweiz
Philipp A. Urlich: www.artstation.com/somartist
0:00 intro
1:01 set 1
2:09 set 2
3:31 set 3
4:58 set 4
6:18 set 5
7:11 set 6
8:34 set 7
9:42 set 8
11:26 final thoughts
12:50 outro

[@VantaDraws]

The lesson I've learnt is that artists need to watermark their works

[@Joeydooodles]

Banana Eggplant Lion Onion Soup Now read the first letter of every word

[@atsushisbf_e]

belos?

[@jmvr]

​@@Joeydooodles Watermelon Eggplant Earth North Orange Rainbow Now _you_ read the first letter

[@p_pthenoob]

@@jmvr wtf is a weenor

[@jmvr]

@@p_pthenoob wiener lmao

[@galaxymoon8071]

I’m genuinely shocked how bad they were at spotting the ai. Like they were all very obvious.

[@insert-name-here6777]

Training. Some just dont see it very often

[@GikamesShadow]

@@insert-name-here6777 bs. Has nothing to do with training. The red haired person got so many wrong just cause they refused to take a second to look at the details closely. Which you have to with AI.

[@insert-name-here6777]

@@GikamesShadow That too. Bur if you know at what to look first you get a lot faster

[@helianthe3457]

Don't forget that this vid was one year ago and it's progressing extremely fast and we're also more educated to it. I also found it pretty easy to spot the AI ones, but I'm an artist in 2024 who's following this very closely. This was just a vid for fun but if we're being real the issue is that I can't imagine most people (meaning non artists, and even artists under a certain skill level + not used to see AI art) being able to spot which is human made and this majorly sucks for a myriad of reasons

[@TheMagicDragon-mm5dr]

That second to last one was NOT obvious, whoever did the left piece... it just looks like AI. Otherwise, yes, nobody should've gotten the second one wrong. The landscape one was also obvious as hell! The architecture just looks like... nothing! There's nothing there! The bottom one is coherent across the board! The architecture on the top, which is the focus of the piece, means nothing, it's just abstract nonsense!

[@Samythecoolkid]

I guessed all of them yay The ai art either looks "melted" for landscapes or they aren't very good at details.

[@The_JINX]

For me, the biggest giveaway for the landscape ones are the background details. An artist knows the amount of detail needed to create an area to focus, while the AI keeps the same amount of detail everywhere, what makes it look more stale.

[@justmilkcat6274]

Same lol

[@fuzzytadpole]

Me too I feel accomplished >:D And yes definitely agree

[@marylit21]

also they're really bad at good light and shading

[@forestjester3738]

REAL GOT ALLL

[@ScriptForever]

I can believe people got the second set wrong💀😭 The beak is wrong, the eye is messed up, and the bird’s arm is just connecting with another bird? HOW DO YOU MESS THIS UP!??

[@probablynotblufur]

yeah i swear the bird’s right eye had no pupil and fur on it 😭💀

[@gayfurryfemboyy]

Plus i don't get how they messed the one after like bruh the hand holding the sword had 4 fingers

[@ItzpoomAV1]

this reminds me of what gdcolon said: Finally, the weapon is clearly made of three components but you only shoot two projectiles. Normally this wouldn't bother me, but his health bar goes down three times! So is there a ghost bullet or something? Well actually, Yes! There is! If you turn off preview mode, you can see that there IS a third bullet, but it stays invisible so you never see it in the level! *HOW DO YOU MESS THAT UP???* **screams violently into pillow**

[@lukaswastaken8756]

My opinion stays the same, I hate it with a passion

[@lorenzomizushal3980]

Easy to hate what you don't understand.

[@user-ik4yr6wc4p]

yeah, not great especially since people make money off of it

[@StrawberrySoda18]

Same ai art is disgusting

[@fazguy4547]

​@lorenzomizushal3980 oh we understand it *fully*, that's why we hate it

[@pubberhubber3462]

​@@lorenzomizushal3980 It's easier to route for something you don't know the downsides of

[@yaboiEchogaming]

Hey guys I would like to give a heads up about watermarks, ai has actually made up watermarks and implemented it into its generated images so don’t trust art based off the watermarks. This goes for stolen art as well. Some people can paint over watermarks and pit their own over it. I’ve never seen someone do this but it’s a genuine fear I have next to someone drawing something problematic and adding my own watermark over it. Be safe out there, stay epic. ✌️

[@compassrose1466]

It’s getting worse because companies are now using ai art Legos, Dc comics, and now even a game Nintendo allowed into their system. So now you can’t even trust actual companies who can AFFORD artists :/ it’s sad you can’t even trust watermarks

[@raidev_]

you can usually tell watermarks made by humans i think

[@Asra_..]

usually watermark=online social media account,if you search the name and the piece is human made you'll find the artist's accounts

[@Verchiel_]

the only possible ways to know if art is AI or not, is either using reverse image search to track the source, which isn't always available. Or learn to detect AI artifacting, which is proving to be really difficult for many people regardless of their skills like being an artist themselves. With every few months passing buy the latter is becoming harder and harder, especially now that AI generated video is becoming a thing.

[@CadaGCreations.]

I’ve actually seen that before and it’s crazy right?!

[@JuhoSprite]

This was a year ago, I could tell every piece that was ai, nowadays I dont think its that easy anymore... im surprised some of these artists here even got fooled by these.

[@moriyamakyon1067]

huh? atm even best ones i could get required so much fixing and still are clearly ai. So no there. They look good but mostly you wont get thing you want

[@JuhoSprite]

@@moriyamakyon1067 depends on what artstyle, there are even videos made by ai that look like real life dude.

[@alfiearmadillo]

@@moriyamakyon1067depending on the style youre going for, its def possible (with a lot of generations) to get 1 or 2 pretty indistinguishable pieces, but yeah if you know what common oddities to look for it tends to be very noticable still

[@moriyamakyon1067]

@@alfiearmadillo actually I generating a lot lately, still best one would require me hours to hide obvious signs of ai, mainly it's lines, colors and blur. It's takes a lot of work to get good results.

[@worm6820]

seriously i was shocked anyone was even questioning themselves on any of these pieces.

[@thatknownali6346]

I have a feeling they will most likely pick out the art on certain details. Like hands, eyes, shading and etc.

[@Mapple318]

I did the same during the vid. Did not expect a character with 6 fingers-

[@outlawsyl]

@@Mapple318 It's not just the number of fingers that AI gets wrong with hands, and in the image the pose is so complex and well drawn to the point that you'd think 6 fingers were intentional

[@lunarul]

number 7 really threw me off. It was too good for AI, but it had so many of the little details that would normally hint at AI that I'm still not sure if the artist didn't either integrate AI into their flow or intentionally produce something that looks AI generated. The hand poses were too good for AI and I was going to overlook the 6 fingered hand, but there were other details that you usually see in AI art. The way that cord ties to a buckle in the middle of the dress, the phone receptor coming out of nowhere with some random tiny mechanical attachment, the wildly asymmetrical boobs...

[@UODZU-P]

@@Mapple318 I fully believe that "artist" is just using AI and passing it off as their work.

[@GamesPlayer1337]

​@@outlawsyl That video is a year old. It became way better with hands btw, not perfect yet but it quite often gives good looking hands as results

[@elizabeth3978]

7:50 the left one has a EYE on the arm

[@binarycat1237]

oh fuck i didn't realize until i saw this comment and turned the brightness up

[@pinkapetdesigns9134]

I saw the weird blob, but didn't piece together that it's an eye! What tipped me off is the hair(?) in the top right corner

[@chupapi7592]

I thought it was a weird fold from the shirt but I could tell it was ai

[@bamer6000]

I thought it was like a bullet wound or something

[@itcanbeverydayforyou]

​@@bamer6000 It does look like it

[@LaysAirchips]

For me it's kind of easy to tell based on the faces, because a lot of the ai generated ones have the same vibe when it comes to the style, especially the face (although you can kind of tell by the way it's coloured, shaded, etc.) Sometimes you can look at the lineart as well, because ai doesn't lean towards having it's own characteristics and variation. Obviously the hands are a good focus point too when it comes to trying to tell the difference, although artists can mess up with the hands as well, just maybe not as much as ai does (it could also be a choice to have less or more than 5 fingers but idk) Also I'm not saying that every artist who has this style and it's characteristics is not an actual artist, those are just the traits that the ai seem to use most in their "art" (especially in NOVEL AI because it does take a lot of inspo from anime styles)

[@rr.studios]

Agreed, ai anime faces look quite sharp and bland in comparison to human artists with similar styles. This was a dead giveaway for all the artworks (the last one has a lot of generative looking features which was a dead giveaway) which made it easy for guess accurately 100% of the time regardless of other lower yield giveaways like strange finger shapes or numbers, eye shape, disfigured clothing etc...

[@thederpydude2088]

​@@rr.studios What about the AI anime girl in set 3? I agree that the other sets' AI faces are kinda bland maybe in the sense that they're less expressive and more neutral, but I feel like set 3 might be an exception. And, since this video is from two years ago I expect the current AI generated faces to be much better, so it's probably even less of a useful tell of it being AI generated. Still, I think AI art continues to be faulty when it comes to the details. If you look close enough, I think there's still almost always something that's off.

[@rr.studios]

@@thederpydude2088 What gave away set 3 to me was contrasting the arms of both, admittedly the faces were not a dead giveaway but the other clues helped a lot. Of course, the trend exists where eventually it will become near-impossible to differentiate between ai and human art. Less than a decade ago, ai 'art' was garbage and easily seen as such. A decade from now, I'm pretty sure no one will tell the difference if ai 'art' generation is trained further.

[@thederpydude2088]

​@@rr.studios Yeah, and I think it'll probably get harder to argue that AI images are soulless or don't have feeling or whatever when people eventually can't even tell the difference in the future. Although, I think people could still have the upper hand when it comes to actually coming up with unique and creative ideas, and they could be able to recognize when something is so good that AI alone could not have achieved it.

[@jeffreymaina6906]

One thing nobody mentions is ai art has a lot of indefinite little tendrils/lines/shapes or arbitrary, non-functional looking accessories or details that just have a mushroom trip or "unable to tell wtf it's meant to be" like quality. It looks kind of like something but not quite like anything... Hand drawn art usually feels more intentional, without those "wtf is that supposed to be exactly" like qualities. Idk if anyone knows what I mean but it's plain to see. Again, talking about small details such as objects in the BG that just don't really serve a purpose--not so much larger details, which AI is decent at.

[@guilhermerafaelzimermann4196]

It's like looking at that one image that feels really nostalgic at a glance but has no defined actual things in it

[@wmoros4902]

the last image is noticable with it, buildings fall from form into abstract lines, and lamps etc that dont make since.

[@thederpydude2088]

Yeah AI hallucinates a lot, not just with generated text like with ChatGPT but with visual information as well. It's just nonsense that might seem at first like it makes sense, but a human who actually understands what they're dealing with would understand their decisions better.

[@lovelydumpling]

Easiest test of my life, but AI has evolved since then and I suppose in late 2022 people weren't as used to seeing all of the AI garbage yet. The six fingers in the Yukuri one almost threw me off but I figured that had to be human error because other details were too human and the one on the right is like that quintessential AI artstyle

[@moriyamakyon1067]

it's like 99% overpaint of AI , literary it's so strange left one has 4 finger on one hand and 6 on second , arms are crippled , dress has some uncanny dark line , idk how she holds pen , phone is literary flying in air and etc.

[@thederpydude2088]

​@@moriyamakyon1067 I was so confused about that actually. A lot of it seems to make sense in the way that a human-made illustration would, but the six fingers on that hand really through me off, so much so that I tried to hunt down the original artwork. I found a place it was posted by seemingly the artist, but there were no indications of whether the artist was one who used AI or not. My guess is that some part of the illustration process involved AI, but I couldn't seem to confirm it. Also apparently the character is supposed to be Megumu from Touhou, but I don't think she's supposed to have six fingers, so, for that illustration at least, AI probably was used. Otherwise I have no idea why they'd draw six fingers. Maybe a mistake, but I'd be a bit surprised if they didn't realize it as they were rendering it and everything.

[@gaymiens]

@@thederpydude2088 extra fingers is up there in the list of most common art mistakes (you can find examples of this even in professional work, usually animation (where each frame needs to be drawn on a tight deadline, so people overlook stuff like that)). you'd be surprised what you can overlook when you're staring at the same image for hours without fresh perspective

[@princembat]

the main thing that told me that the yukuri one was real was because of the background, because theres actually a phone stand behind the character, and shes talking on the phone lol

[@codemanthe2nd343]

The set 8 landscape was so easy, I can spot stable diffusion nonsense "detail" from a mile away. I've never seen an artist imply detail like that, the lack of intention and procedural noise is really ugly when you look at it closely. The overall color choice and lighting was pretty good though.

[@INTHISSILENCEIBELIVEEEEE]

i got all of them right! heres what helped me! in number one with the blue haired girl, (obviously) the hands sprouted more fingers and whatnot. in number two for the bird, (obviously) its beak is inverted. in number 3 a blond girls sword case fuses with a lighting bolt in the backround, one of her hands are also missing. in number 4 the blond girls ai tries to sprout another iris within the exact same iris, leaving a quarter of an eye just hangin in there. in number 5 mikus leg just has no joints, curling like it has no bones, muscle or anything. as if made out of elastic in number 6 the guys iris comes completely seeps out of the eye, and has no pupil either, not to mention the hair fusing with the corner. in number 7 the womans eye whites just completely fade into her skin. also, her iris are trying to duplicate just like the blond girls eyes. in number 8 a boat fuses with the stairs\dock or whatever it is, and the roof duplicates and smushes togeather, just like the wiring on the sides.

[@Kodaiva]

it also helps 3 has 3 fingers and a thumb on the visible hand

[@lmnpie2]

I can immediately tell by the eyes. The eyes always give it away.

[@repeatedecho]

Got all of them correct. It was pretty easy to tell which one was AI(the last one was pretty difficult though) based on several factors: 1. Hands 2. Artstyle 3. Facial expression 4. Rendering and shading of the hair, skin, clothes etc 5. Background detail Hands is pretty obvious, Ai art cannot do hands very well. They are often hidden or turned into some sort of eldritch abomination. Ai-generated art has a certain artstyle to it. Its especially prevalent in the face by the way the way the hair is highlighted and rendered overal. Scrolling through ai generated art can help you better recognise its artstyle. Take into account which ai program was used and look at the differences between images produced by stable diffusion and images made by NovelAI. Facial expression is also another way to distinguish ai art. Characters made through ai art tend to have very little emotion in their faces. Rendering is also another dead giveaway. Mid journey pieces are very colourful and full of smudgy details. NovelAi shading and rendering is usually very simple when it comes to the hair, but VERY detailed and pronounced in other areas like the chest. The rendering of clothes in ai can have this smudgy look to them if you look really close. Background details in ai art can also have this smudgy look to them.

[@Aquilenne]

But... in the second to last one it was the real one that had the super screwed up hands

[@ttvflakitgh]

Ai art always has a very distinct shading and generation style. Most artists should be able to identify it at a glance.. it’s not difficult.

[@PancakeLeBear]

5:01 the art piece on the left really gave it away for me, since there is a signature written in Japanese. Overall this challenge was pretty easy for me :)

[@Flyingsearat]

I’ve also noticed that AI struggles to write , ESPECIALLY in scripts that aren’t the classic English (?) one 😭

[@TheAroSpecWallet]

Wait where was the signature exactly?

[@BlueBonnyAnimator]

@@TheAroSpecWallet i think they meants the symbols(kanji) in top right corner, what i assume is surname and name of an artist

[@aminakay661]

Artist here! HOW DID THEY GET THESE WRONG?! is it really so hard to some? And no it has no place in the art community BeCAuSE it was not made with the intention of it being a tool for artists. The official statement says it was but I don’t buy it. If they really thought about artists they would have compensated them and asked for permission. They only care for the profit and nothing else and now we can’t trust anyone.

[@helmikusu5609]

I guessed all of them correctly just at first glance. I have enough experience with both art in different kinds of fields and AI art as well. What was mentioned by more people in the video, the feeling that something is off or that the AI art is emotionless is actually not a coincidence. From the beginning of times art was used as a way to communicate information and share ideas. this is something that is retained even nowadays. When an artist draws anything it always has a purpose, even if subconsciously at this point. Art is also different with your own personality, how much you learned about whatever you're drawing as well (the amount of knowledge changes the overall visual of the subject) Individual artists have their own series of shortcuts and their own interpretation of reality that they use in their art. These "simplifications", shortcuts, interpretations... this is what we call an art style. It is individual to each person and it is affected by the way we see the world and who inspired us. This is what AI can't do. It cannot communicate ideas. It's like when it tries to write text in an art piece. That is how it's trying to communicate to us what it's generating as it is white literally an amalgamation of all artists and it doesn't create with a plan or purpose in mind, it doesn't want to communicate to us, it just calculates the likely probability of what culmination of all artists could create. however it is not filtered by an individual and their idea and the actual "style" which is why the art appears soulless, confusing, almost generic. Theoretically you could say AI is like an artist that took inspiration from every single one artist in the world and attempted to use all art styles at once. it will simply not work. It's also why eyes are often muddy and messy because "eyes are windows to a soul" is something, we humans, tend to put a lot of emphasis on and every single one of us has their own unique take on how we draw eyes. This confused the AI and they can only average what a typical anime eye looks like but they struggle with any other eye type because the amount of inconsistent data it received from all artists combined messes with the result. This is why fingers are often messed up because they can be in so many different positions and you know there is 5 fingers on a human hand, but AI doesn't think about it and they only have thousands of inconsistent images they have to kind of "average out" to create the hand. It can also result in confusing hand gestures which do not communicate anything to us. This is also why snakes are almost impossible for AI to draw as well because a long legless noodle that can be in all kinds of shape or forms with a head and beady eyes is too confusing for the AI. They don't think, they only take the most average result based on how you limit them with your prompt. And the more you limit them, the worse the result as they have less and less examples to work with, unless you specifically provide them. I think AI, as intimidating as it is atm, it is still not perfect and I do not think it will be any better any time soon unless they give the AI an ability to think and learn like a human does, processing logical information and attaching visuals to it. Until then it will just be the average static amalgamation of all artists. If you ever feel like ai can do better then draw a danger noodle, hands, or your oc being expressive with a meaningful background, interacting with someone else or an animal companion. *Because AI could never.* that is all great vid! 😂

[@simplicityd8703]

Well AI art has gotten into scripting and now I'm getting scared

[@Michael-Humphrey]

it already writes and drives drives to O AND controls all the social media algorithms

[@simplicityd8703]

@@Michael-Humphrey i only use youtube but the fact that it can do so much is getting scary once it learns to model anyone will be able to make a game

[@jadamiller7485]

I saw the cat made fuzzy in the rain, and knew it was real. AI will never ever do that.

[@frenkieshis]

I got ‘em right cuz AI has a really specific “artstyle”

[@SL33PY_Gacha]

Same, I don't know how I would describe that "artstyle" but you can just tell immediately

[@frenkieshis]

@@SL33PY_Gacha yeaa its not an artstyle but IDK BUT AI STUFF WAS/IS ALL THE SAAAAMEE

[@SL33PY_Gacha]

@@frenkieshis fr I always call "AI art" AMI artificially made image because it's not art so I don't wanna call it that lol

[@personalalt317]

It always looks so smudgy

[@FyreMoon1]

@@ItsLeBridgeand the asymmetrical boobs, the random tower thing in the background, i am 100% sure that both of the ones on #7 are AI, just one of them is someone faking real art

[@StarFried]

I only got fooled by set 7 and i think it was because the blue one had so many details that i barely saw the other one

[@anxia-tea5846]

this is actually a very good way to test and improve your observational skills

[@mcDoYanfei]

5:34 immediately thought the 1st one was hand drawn because of the chinese letering

[@LaurenAlas]

I think it’s Japanese, as a half-Japanese person, but tbh they share kanji so I don’t know if they have something like this in common or something

[@mcDoYanfei]

@@LaurenAlasyes kanji and chinese (simplified) do have some similarities but i can decipher the words being 山问菜 and 木

[@pizzaaaaaa36]

6:24 WAIT I REMEMBER THAT ARTWORK (the handdrawn one)

[@Abigail-mi6ib]

11:39 oh it can now and that worries me :/

[@moonegalactica]

5:19 fox ears has a clear watermark, that you can tell what it says. 山河草木 in the top corner

[@shiny2575]

If you know where to look and what to look for, its really easy to spot the real and the fake. Real artworks tend to have visible brush strokes (usually with a round brush), cohesiveness (background has a tree, and there are leaves floating in the wind) and when theres a character the background tells a story or lights the character with its colors. Ai art tends to contradict itself with the general motion of the piece (while one part of the character looks like its going one way, the other goes another way, and it looks like theyre doing an unnatural motion), blend in the details and generally not tell a story with its elements. Thats why the more realistic ones are just some anime looking women standing around doing nothing

[@ToastTheGhost1]

I got all of them using a method I usually use: 1) does it have writing? Numbers, signatures, words, they all help because AI really struggles with it. 2) are there any broken lines or “anomalies”? E.g. an extra finger. Number 1 overrides this if it has correct writing though. 3) how many art styles are used? Is the background one and the character another? This is a telltale sign. 4) this depends on the options really and is overridden by the other signs as it is a style: AI, when drawing anime, has a very specific style that it strays towards. E.g the Miku one, the blonde one with the star, the blue hair girl, the blonde girl wearing Green and the black haired girl. It is still an art style, but AI finds it easier or something and uses it a lot. Hope this ever helps in future even though it probably won’t :)

[@ASharkNamedWaffle]

How can so many people guess right on the second to last one? The blue girl has messed up hands (like an AI does) and yet that one was the hand drawned.

[@ToastTheGhost1]

@@ASharkNamedWaffleI got it because of the clock, it has correct and not-warped numbers (which AI can’t do) But you are correct, it has a very AI-like style

[@MrAfrodisiakOo]

@@ASharkNamedWaffle the second image in that round has a very distinct AI art style. Also melted irises, non-sensical random highlights on the hair and very liquified looking torso. Oh, and also the folds on the dress are weird. Artists can draw messed up hands and make mistakes, but it's usually different from how AI does it. Also the character can canonically have 6 fingers. Having an abnormal amount of fingies isn't the tell for AI, it's when those fingers are drawn super weird like they melt together.

[@GamesPlayer1337]

​@@ToastTheGhost1couldnt do a year ago* There are some AIs that can put numbers and writing correctly sometimes. Yes it is just sometimes, for now. But we know how quickly it can evolve and once the models are good enough and can be stitched together so one model takes numbers and writing while another one does the drawing and so on... You wont be able to tell anymore from numbers. Some models even draw hands now somewhat consistently with good results

[@ToastTheGhost1]

@@GamesPlayer1337 What.. Okay, so science needs to pause, take a few steps back and go down a different path, because we clearly messed up somewhere. What is the point of having AI which has the ability to do that? Smh 🤦‍♀️

[@milkywayssxx]

3:37 “I thought the bottom one was from fate” the fate character right next to it:

[@lttrickster5643]

I think the problem is what they’re looking at- instead of looking at mistakes you would make, look at it in a way that a non-artist would make (it’s being trained by the work of artists, but keywords are given by non-artists). The thing that gives it away most for me are poor design choices (clothes that don’t have seams or a dock for boats that doesn’t have anywhere to leave your boat), when weird mistakes look like something else (like when the back of the buildings at the end looked like hair- the ai started adding hair to the image probably because of “dark” being a keyword for dark architecture and dark hair) and mistakes in the complex anatomy- ai doesn’t just add an extra finger, it adds an extra finger with two joints, while an artist knows how to draw fingers but may not take the effort to count how many fingers are on a hand if it’s a long piece.

[@CheshireGhost]

I wish people would call them "AI generated images" or "AI images." Because that's what they are. They are not art by technicality or definition.

[@thederpydude2088]

Personally I do call them AI images when I'm not looking for people to argue with me about whether it's art or not, but I don't think it's fair to completely disqualify AI images from being allowed to be considered art. Sometimes you might call it art just because it's a pretty picture. But I feel like the most concerning thing is that, by saying it's not art, people might suggest that it has no value. Memes are art in a way, right? .-. I think the enhanced meme creation potential of AI alone might be reason enough to consider that it might actually be appropriate to call it art.

[@Theultramadman]

Bro you can totally call it 'AI Art' since the full meaning is Artificial Intelligence: art, which fully describes its meaning. However if you meant by originality and actual art, then you're right.

[@thederpydude2088]

Let's gooo B) AI still can't fool me. Here are the things that stood out to me: 1) 1:01 Yeah. Hands. 2) 2:09 I feel like the hand, or whatever the small bird is supposed to be sitting on, is a big giveaway. There are a number of other things, like the mildly nightmarish eye and the weird shape of the container, but I feel like there will obviously be some things apparently wrong, confusing, or just weird the more you look at the image. 3) 3:31 The hand in the right image quickly gave me the impression of an AI generated image. It having 4 fingers could be a stylistic choice, but that's less common than AI simply messing that up. Also, if a human drew that, I think it would be more clear what on Earth those red shapes are supposed to be. It's like a sword but also a part of her outfit and also not matching a ton with the top of the sword. The character also has inconsistent shapes for her pupils. 4) 4:58 Blue dress girl's hair on the left seems to have a few floating bits. Of course, this could be just the style, but it also seems random with it only being like that in that one spot. There's also definitely some weird stuff going on around her neck. Her hair and outfit are melting together 🥴 5) 6:18 For set 5, I jumped to the eyes and hands again (AI seems to struggle with them both), and I thought the apparently disproportionate shape of the hand in the right image would be an easy giveaway, but I think that was just a human mistake. What is kind of a giveaway for the left one is that there are kinda messy imperfections in the eyes, and it was also odd that a clump of hair on the right side of the image seemed to be fading into the background despite being apparently the same depth as the rest of the hair. Oh also, the hair near her face on the left seems to be doing something weird. It's affecting the shape of her head and its curve under her chin seems a little broken. 6) 7:11 (Haha Seven Eleven... Anyways 👀) I got a strong first impression of the right artwork being real due to how many visual elements seemed to be balanced well. There's a lot of soft and hard edges and textures and effects that I feel like an AI would struggle with replicating. On the left, the weird hand was a bit of a giveaway when Skynix's avatar wasn't blocking it anymore, but before that I noticed the shape of supposedly the hair in the top left corner. It seemed like AI because I was struggling to tell what was supposed to be going on and it seemed like the AI was confusing the visual elements of the hair with some other thing it was hallucinating. 7) 8:34 This was the hardest for me. Seeing the six fingers and that seemingly random shape in the background of the left image really through me off as something AI might generate, but of course it is the real deal. Everything else seemed so perfect and didn't have any weird inconsistencies. For the right image which was made by AI, it was tricky to find any obvious giveaway due to it being a simple image overall and not having many elements that could be messed up, but I did notice some things. The eye on the right has a weird bump in the iris, the patterns on the dress on her shoulder seem like visual gibberish, and I'm realizing now that the hair also doesn't make sense. There's a strand that combines with another and a small strand that seems to jut out of another strand randomly at a sharp angle. 8) 9:42 I actually found this pretty easy because in like a second I noticed the top left corner of the top image where there's just a jumble of pipes or something except the AI was so inconsistent with the thickness and structure of it. Also in the far left center there was this odd foreground lamp structure that seemed to be melding with the background. *Some things you might take away from this:* I feel like a really big fault with AI is that visual elements that logically aren't related can often be found affecting each other. Stuff like objects in the foreground or just that would be thought to have some physical distance from something behind them will affect those background things. Also, AI can be bad at keeping the shapes of edges or lines consistent. Is this line supposed to have a smooth curve? Is it supposed to be straight or converging with another line? Usually there are some parts of the image where AI gets indecisive about this and changes its mind about the line half way through. AI also kinda sucks at patterns. If there's any repetition that you might expect to be consistent, check to see if it is because often there's something inconsistent about how AI depicts patterns. Same thing applies to symmetry. Pay attention to when you see something unexpected or confusing. I often look for mistakes in the eyes, hands, hair, or clothing. Those seem to often be problematic visual areas. AI's mistakes are hardly limited to just those though, so keep an eye out.

[@fennwenn3317]

Since there's so much miku artwork floating around the internet, AI is really good at rendering her.

[@mattie5111]

I got set 7 right by looking at the more complicated details (like the wavy hair) on the blue image and seeing that they made sense. AI always either sticks to really simple hair (like the second image) OR tries to do something more complicated and fucks up the details so seeing something complicated with details done right is always a 'that's a real artist' for me. it worked on the final set too, the architecture painting i IMMEDIATELY knew was fake as soon as it was on screen because some of the details made no goddamn sense

[@ut3115]

arg fail only on the 6 fingers's one!! i wasn't ready for this

[@Memecats69]

I love how you support the art community by implementing smaller artists in your vids :)

[@petre1758]

Hands on the human art in 4th and 5th set threw me off, I guessed the 7th one initially, but doubghted myself into changing the answer when the gests started speaking. Got every other set rather easily so that's good. nice video

[@sirinslament]

I’m an artist. Got all of them right. You can notice the tiny details of actual rendering, when line art is clipping into itself, etc.

[@PJBgamer]

For me, best case scenario for AI art is that it's used as a tool to inspire someone rather than going "hey I used this AI to generate a piece and it's totally my art." Of course, the training for these AIs should all come from art that was given with consent. Also, that last one really stumped me, so I was utterly blown away to see Seb spot that teeny tiny signature. This also makes me wonder what will happen if sentient robots start making their own art... Is that AI art?

[@Ruth25853]

well, for backgorunds you can tell its ai when there's no structure, specially for man made stuff, if it makes a library for example, you'll notice the ai cant make sense of the small props, like, it doesn't understand a book is a book. A human artist, even if they simplify those props, they don't make them 100% abstract. With a close up of ai you can actually tell the material is just no good to use directly for production because it makes no sense

[@Mayiamaru]

The biggest give away for me spotting anime AI, it's always the bangs for me. They for the most part have the same type of fringe. Next is hands, eyes, then other details.

[@guiiada]

Ai should be morally developed without stolen art, they should replace jobs that are dangerous or no body wants to do. they should be a tool to help artists not threaten them

[@GamesPlayer1337]

There are artmodels without, what you call, "stolen art" btw

[@BlankedOutAce]

though, even without stolen art, i would most likely never give a compliment to an image i know took 3 seconds to make like, if i saw a background on a game, and knew it was AI generated, i would never go "damn, props to the devs on this one"

[@thederpydude2088]

​@@BlankedOutAce Maybe the technical expertise that allowed AI to be created in the first place is commendable, but yeah in most cases I've seen, images made with AI aren't really a testament to the prompter's skill or merit in any apparent way.

[@guiiada]

@@GamesPlayer1337 whats with the "stolen art" quote?

[@GamesPlayer1337]

@@guiiada Because the "stolen art" is atm of writing this not stolen as defined by law in most, if not all, countries. If the law changes and deems it illegal i wouldnt use quotes, but as it stands its legal and a gray area at worst. Theres ongoing lawsuits but nothing with a real conclusion yet as far as i know.

[@Nabit808]

On set 2 the bird holding a bird just has a human hand that becomes the smaller bird

[@yaboiEchogaming]

Also the eye seems to spiral into the eyelid The hat doesn’t seem to be finished The beak is inverted The bag hook(?) has a strange vanishing point The tree has a random area that seems out of place And the left iris is missing

[@wanna_be_monke]

keyyuui is a menace! no way she/he draws hands so ai like

[@Aki_The_Weirdo]

my opinion has not changed, I don't want it anywhere near my silly art and AI basically steals art since it like combines other art and tries to make it, but it looks so bad. Also AI's pretty easy to tell from actual art.

[@cottoncandyperfume]

I do have to disagree with this whole stealing thing. Artists also get inspired by a lot of their favourite artists, and grab parts from them to put into their own style. I'm still mad about AI art though, feels like my whole carreer is thrown out for an emotionless machine.

[@silverdirtdraws9828]

​@@cottoncandyperfume AI isn't interpreting the way people do tho. If I take inspiration I'm interpreting what I'm seeing and using it within my own style. AI just takes directly and does a mash which is why it struggles so hard on the small details. Because it's not interpreting it's just mashing

[@septiquaddoubleyou4019]

@@silverdirtdraws9828 there are real artists who an struggle on the small details as well. I've seen plenty of amateur artists whose styles are just a mashup of stuff they saw elsewhere.

[@tyrnanreply958]

agreed also art is supposed to portray the human experience which AI art obviously doesn't do like regardless of if it looks good or if its stealing art it fundamentally misses the point of art cause I don't care if standard anime girl looks good if its drawn by AI cause it isn't attempting to communicate anything the way art is intended to its just crashing good art together and creating something without meaning

[@Chris-Clips-Games]

@@tyrnanreply958right, as much as i am pro-ai, ai “art” isn’t art, its just output. I don’t think Ai Output should be supported because its just grabbing different art and mashing them together into gross mash potatoes.

[@kyoto5463]

Got them all right, AI has a very specific anime art style, and even if it doesn't have it the details instantly give it away (which is how I got the last one, powerline in the background was messed up)

[@ASharkNamedWaffle]

9:31 The left one has to be an Intentional fake AI. Her right hand has a wery long finger while her left hand has 6 fingers.

[@IceDiamondArmy]

yah i notice that too. I think there might be an error in the video.

[@squiddler7731]

I also noticed that lol. My best guess is that the AI art was good enough that it fooled someone in editing and they got the pictures mixed up: no way a real artist accidentally gives a character 6 fingers

[@ASharkNamedWaffle]

@@squiddler7731 Unless the character actualy have 6 fingers. How they are holding the paper makes it feel a bit weird. And the pen looks like it is floating. (but you can do that effect by holding it with your thumb)

[@Bryan-yv7kf]

Yeah there's so many little details and errors that gives it away that something wrong, i was even thinking that both of the drawings where Ai and put side to side just to mess with everyone because "only one of them is real" but no, ended up with that being "the real one" and idk there's something about it that it just rubs me the wrong way

[@hulfe2514]

Set 7 where the girl in a blue dress has 6 fingers threw me off so hard

[@eliclary2318]

Crazy to see this now and how far AI has come, it can even animate now lmao

[@kraykat4514]

the detail that everyone overlooked it's the complexity of the drawing and the LIGHTING , no single soul ever look at the lighting are roughly the same and just focus on the hand is drawn correctly or not, that's why a lot of them are fail at set 7

[@KaedeBidoof]

this my first time seeing a video with the new vtuber model, *exquisite* 👌

[@Bloodanna]

Got all of them except the last one. The last one I was really torn because the building so obviously had parts based off of Dishonoured Fanart I have really loved in the past that I didn't notice some of the more obvious problems.

[@lapiswolf2780]

I didn't notice the hands on the first AI image. I told it by the shine on the hair.

[@tylove7992]

Got every single one right. AI just looks too perfect. There's just no soul behind it.

[@GCatlord]

IDK how they thought the first AI landscape was "good" it was hellish to look at for me. So much scrambled nothing blending into more nonsense details. It's like looking at a bucket of writhing ants, it's awful. Also too many people used the hands to discern AI art from real art, sometimes it nails the hands, or just hides them. Sometimes real people make the hands weird, like miss 6 fingers over there! You wanna know the dead giveaway for that one? The clock in the background had real numbers, in order. 4, 5, 6, 7. The inkpot and plant in the background was visibly an inkpot and plant, and not blobs. The phone she was holding didn't blend into her face or clothes. Those details really show what the AI can't copy.

[@hercar20xx5]

For the second one one thing that tipped me off was quite a few ai’s have a weird obsession with that type of orange glow lighting but the you can also tell from the messed up hand and eye

[@jstiller30]

Got them all. However, the 7th was somewhat of a guess for me. Ultimately the clarity small details at varying depths like writing and the quill/ink container made me decide it was likely human made.

[@RewRewster]

I got an 6/8! Most of what I used to identify the AI one's was because small details.. For some it was very hard cause nothing stood out from the AI one's compared to the Hand-Drawn ones (But I did notice a lot of mishaps when it came to the hands and eye pupils and other shapes). Most of the art displayed looked super pretty and I payed close attention to the details as well as form so it was very shocking when I had reason to believe one of the other was AI but wasn't. Overall from this experience I think my perspective on AI art next to IRL People art is that I usually assume that more people who make pretty art usually put a lot of thought and concentration into their art into making sure it looks right (and it does look amazing, but a couple chosen to be put next to a good AI drawing tripped me out when determining which was generated).

[@SquarahKat]

I often see AI art in the wild, aka on presentations, ads, or it youtube videos and always point it out to the people around me

[@superiordweeb5725]

ai art has gotten way better recently, i'd love to see a revisit of this idea!

[@DinoDoodlesS]

i paused before they spoke and my guesses were all right!! (i haye ai with a burning passion)

[@Lapix-.]

All except for set 7 due to the 6 fingers on her left hand. My first thought was left being ai but I decided I should try to look closer and I just couldn't see a world in which someone draws 6 fingers on their art 😭

[@edgalice]

Perfect score here. AI doesn't know how to deal with a line that goes behing something else (Landscape one: on the left behind the light, the building disappears ; Miku one : her left jaw (our left) his weird)

[@kirbydswii]

I would like to see a new video on this as it would be harder if it was just one image. Some artists have art styles similar to ai art so just having an image and guessing by itself would be harder. With two images it is easier since even if you are not sure if an image is ai, you can be more sure by comparing art styles.

[@jademonass2954]

maybe i just got a keen sense, but i got all of them right haha they have these telltale signs i cant describe

[@chaoskonig8755]

I could tell all of them apart, the AI Art just feels soulless on closer inspection

[@stephaneedward4367]

I guess all of them almost didn't notice the mess in the top left corner of the landscape one until the zoom in

[@ThedoorMan-zh9ln]

One of the is an amazingly crafted piece of art and the other makes my brain hurt if I stare at it for too long.

[@Akaykimuy]

I honestly don't understand how anyone would get the human ones wrong

[@AGayPersonThing]

Ai art has a very distinct warping

[@Your-Average-Nerd]

9:45 On the top picture, you can see a reflection from the sun/moon in the water even though the building would be blocking it

[@Taiyunn]

The way I got them all right so quickly LMAO

[@Markty07]

9:30 I think the 6 fingers are either on purpose or artist mistake, but that clock is just too perfect for AI that's why despite the 6 fingers I'm still picking real.

[@FyreMoon1]

I think the artist of the left faked it, and used AI 1. 4 fingers on one hand 2. 6 fingers on the other 3. How the hell is she holding the pen 4. Why does the wire x just vanish 5. That weird structure in the background that doesnt look like anything 6. Asymmetrical boobs

[@princembat]

as the great marina says, "you may be good looking but youre not a piece of art". also as an internet user said that i forgot the name of "why should i care enough to behold something no one cared enough to actually make?" i got all of them right, because its all in the details. also, ai images tend to have a melting feeling to their details, especially like in that last one with the weird sea house mansion thing. like, at a glance these things look like something, but you look closer and its like... what even is that? like, what am i even looking at? is that meant to be a lantern on the left? why does it look like it should be on the shore, but also right next to the viewpoint? why isnt the lantern part centered on its stand? no my opinions definitely did not change lol

[@WierdArtistMarcell]

11:34 *Curb your enthusiasm starts playing*

[@thederpydude2088]

Sora 👀

[@Its-a-me_Z]

I got all except the last one, didn’t see the watermark and thought ai wouldn’t be able to do architecture I agree with all the artists’ opinions on ai, can be used to advantage but shouldn’t be completely depended on

[@jelloburger]

To explain the option 1/2s switching and top/bottom switching, I think he is sending them images on discord and they are giving an order off of that. Just clearing that up for anyone confused

[@plushehchan8643]

This was a really interesting video to me, because there were a lot of points brought up by other commenters (hands, signature, overall lack of life) that point out whether something is ai or not, but one of the things that really stood out to me was this: AI lacks theme and atmosphere. The AI pieces presented in this video, ESPECIALLY the Galaxy/star girl one and the yellow post-apocalyptic pieces, were simply sticking things together, and for the art level the style suggests, a living artist would likely take a step back and try to incorporate them better, ie. the kitsune was surrounded by a very consistent but simple background, a reappearing rope and prayer slip theme, or the woman in the rain being followed by cats, whose silhouettes were realistically distorted by the rainfall. Artists who are generally early to the craft may fall into this trap as well, with a background/surrounding not really fitting the character, in fact there are many instances in the old art on my channel (all of it), but it is clear that the artist is simply new to it, with wonky anatomy or clashing colors or just things that look a bit off/bad, suited to their experience. AI art is generally quite polished, meaning that any disparity between motifs and focus puts a sense of uncertainty in our minds, even if we don’t quite know it. It’s like taking different puzzles with the same cut and piecing them together- they fit, but they don’t look right. AI art lacks purpose in its creation- it doesn’t have a vision for a desired outcome, only a possible interpretation of words based on existing material. 11 y/o me would have pitched an absolute fit if her Blame It On The Rain warrior cats ‘amv’ didn’t look JUST RIGHT, ai will never top the unrivaled passion of a preteen who just learned she can draw cats in ms paint 😔✌️

[@vivi_needssleep]

Fortunately (or unfortunately), I was able to guess all fo them correct! If you look close enough you can see how the AI art always has some deformity and, most of the time, has trouble with consistency so pointing those issues out makes it quite easy to tell. Either way, very informative video :]

[@Stratelier]

I managed to get 4 of them right (1, 2, 3, and 7). Note that watermarks aren't necessarily proof enough, because (a) some generators may stamp a generator watermark on the result, and (b) some watermarks actually get picked up from the training data! (Which usually renders them illegible but they will still "look like" a signature)

[@SailorSketches]

Really Fun Video!

[@azuraslegs]

The eyes. Always the eyes. Also, wrinkles in flesh or fabrics. Also also, when two objects have similar colors, the ai gets confused. For example, look at the AI image of the blond girl holding the pink sword thing. Her left arm (on our right) disappears into her other one. Also the sword itself fades into another object rather nonsensically. The fastest way to tell for me is eyes tho, compare each eye. If the pic is unedited, they wont match in the right ways.

[@yourdad3275]

with AI art many details lack any discernible intention, with the real art it’s clearly deliberate and you can tell what the artist was trying to accomplish

[@agoodmango4124]

The landscape one was easy, the buildings suddenly bent and the land became water, there was random little roofs

[@Lost1207]

For last one, no AI art generator would be able to make that good looking of a dragon silhouette

[@samamies88]

youtube kept recommeding me this video and for a week or so i didn't click it because i was expecting more bias and extremism - glad i gave it a try

[@Khalis2]

I' love , when you using hopes and dreams . ( I'm an undertale fan )😅

[@soulatte8902]

I'm strangely proud I got most of these right, lol, except for one the bird one in set 2 was what tripped me up XD the one on the left was just staring into my soul

[@TheCottonCandie]

the only one i tripped up on a little was set 7, where i was fully ready to say they were *BOTH* AI, as i recognized the second one's AI model used (Novel AI was clearly the one that made that), however, Yakuri's art... it felt human, for sure, but then your friends made me start to question myself and waver, rip.

[@nowishka]

My brain has been trained to identify AI art at first glance thanks to pinterest being flooded with it 😭

[@oM477o]

I got all of them. #1 wonky hands. #2 messed up eye and missing hand holding bird. #3 very messed up sword sheath. #4 the real image has text in the top right. #5 The edge details in the artist image is difficult to replicate in stable diffusion. The character style for miku is also much more common in anime sd models than the artists style on the right. #6 almost got me. Ai generators like to make portraits like the artists picture but the artist image has lots of details like the hair pin that an ai would struggle to get correct. #7 an ai would mess up that clock on the left image. The right image also uses a style common in ai models. #8 it would be damn near impossible to prompt current ai models to draw a distant dragon flying in the clouds. The high dynamic range in lighting accross that image is also something ai would struggle with. The top image on the other hand has weird lamp posts and the whole building looks like an ai fever dream.

[@-ImAb0zOxoxo.]

all these threads on twitter really helped spot AI art…

[@anddatkat796]

I kept thinking "How could they not see that x looked so off" every time anyone got it wrong, but then I realized that the video is a year old and while AI has unfortunately improved further, so has the average person's ability to spot oddities in AI "art". If these artists took this quiz with the exact same examples today, they probably would get full points.

[@The_Kino_Reyy]

I got all of these correct. Not because I have a good eye for art, and im most definitely not good at pointing things out, but because I’m very well versed in getting AI art done for me if I’m thinking of a character for my stories. heres what ive observed : 1 : Very obviously, her hand is messed up and another thing is that AI usually go for the more colorful plain characters while artists tend to use duller colors a lot more while also added a lot of detail to the character. 2 : Sometimes the AI messes up and makes the eyes look like they just swirled a bunch of colors together and called it quits, also the hand mixing in with the bird is most definitely not helping. 3 : At first, I thought it would be the red dressed girl because of the way her hand is positioned, but taking a closer look at the blondie, I could very easily tell that the way the sword was shaped and made was very odd and she was mixing a hand and a few fingers. 4 : This one was a dead give away to me, for some odd reason, AI can not understand the concept of words or letters going into an art piece, so the writing on the first one was a nice hint that it was hand drawn, also the star in her outfit was like melting, very odd. 5 : Short and simple with this one, Ai tends to stay away from adding items or more complicated features like the bag or mark thing of her face. 6 : The background is very confusing, you don’t really know where she is or what she’s doing, some of the things in her hair look like metal sticking out. 7 : I can see why people would be confused on this one but the Ai one was just to plain to be hand drawn 8 : Omg I have a VERY personal issue with anything involving Ai backgrounds and you might not be interested but I rlly do not care, backgrounds made by Ai tend to be very similar and very complicated for no reason. I could say give me a mounting with snow and it will show me a Birds Eye view of A MOUNTAIN RANGE WITH SNOW TREES AND CLIFFS. IT DOESNT EVEN LOOK GOOD IT JUST LOOKS RUSHED AND MESSY. THAT OR RUSHED AND SUPER PLAINNNNN

[@neonnora4346]

I guessed all right The best way to recognize ai art is the weird smudges, inconsistencies and colors in places that dont make sense Also ai really seems to struggle with patterns, they often look smudged

[@yutamo288]

I got all of them correct I'll write down my reasonings, with some observation you can really tell what's AI and what's real. 1:46 - The photo on the left has messy/merged fingers which is an obvious AI error. 2:34 - The bird on the right is not even standing on his hand, its trying to make a cap but decided to use hair to finish it, the beak on the large bird is messy and the eye too. 3:42 - The character on the right has the sword kind of fusing with whatever that thing is on the bottom, and the sword is a little bendy, the hand holding the sword is backwards. 5:34 - The real artist is on the left because the perfect signature and rope on the bottom left being blurred, which only meant the one on the right must be AI. 6:43 - The one on the left is in a different outfit because AI is not good at making accurate outfits so they decided to put a red dress on so you cant compare it unlike the original outfit. 7:35 - The one on the left has a weird thing on the bottom, looking like its trying to make a hand and eye there. I didn't see the fingers because the avatar is in the way but fingers giveaway too. 8:50 - The real artist is on the left because the background details make sense rather than a simple one like the right, AI is bad at making text or numbers and you can see the clock has a clear 6. 10:01 - You can tell the top one is AI because its hard to understand some details, like on the top left is some kind of wires just exist to end. Some spots also looked kind of garbled to me. I'm not an artist by the way. I'm trying to be one but progress takes time. I can tell you, it's worth learning how to make good art compared to generating it. Because at least you are in full control of the results and details included in it, and you'll have a unique artstyle.

[@GreyTrafficCone]

I guessed all of them right almost immediately. Some are dead obvious, others you have to look for details. You gotta think "would it make sense for a human to draw this?" Or "would an AI have an understanding of how a clock works?" (8:34)

[@theredcupcake]

my guesses: 1 - left one (messed up hands) 2 - right one (the tree looks off) 3 - right one (ai, the girl has 4 fingers) 4 - left one (the difference in styles from the background to person got me) ❌ 5 - right one (her hand threw me off) ❌ 6 - left one (weird fingers?) 7 - left one (she has 6 fingers now omg mutation) ❌ 8 - left one (the lantern on the left is a bit crooked)