hi veno. I'm an illustrator myself (leobrunogkk on Artstation) and I would like to give you some feedback on this video, and express some concerns. First of all I think it's a very good idea and it's important to have these videos. Some of the warning signs here are great, unfortunately i think others may very much cause confusion and put both beginners and pros in the targets of people who take your advice the wrong way. This post will be long, so sorry if it gets boring, but i can't think of a better way to structure it. The opening section about looking at the gallery is definitely the most infalible method you suggested. I would like to add to your "Subject matter" and "sheer volume" the amount of styles and how neatly chronologically layed out they are. A working artist, beginner or professional, even if they work in many styles, will generally either have a clear progression or have a scattering of styles and colors through time. An AI prompter will have neatly repeated blocks of identical color and style and subject matter in their galleries. Easy tell. The "AI Spew" section is also definitely spot on. When it comes to intricate detail it does devolve into nonsense when the AI does it, but I would like to make a distinction in this "nonsense". A lot of artists who are not working in a professional capacity, or are posting works in progress or speedpaints will have unfinished details. Not every artist is master of mechanics, or foliage, or other types of detail, so I think it isn't responsible to say "If this detail doesn't make sense, it might be AI". Until we're at a certain level, a lot of what we draw "doesn't make sense". Instead, there are very specific noise patterns that AI tends to recreate, and you glide through a few of them during your video. One of these patterns is the "soft-scaley" noise that can be seen in the Elf character's white undershirt. In the same frame it can be seen on leaves in the background, the edges of her leather armor, and other places. A different noise pattern that is recurring is an almost HR Giger spine-like repeating pattern that can often be seen in more "realistic" images (it's very visible in those recent MTGA promos). Another noise pattern that is in basically every AI image is the "Oil Paint" effect, which occurs from the algorithm each model uses to denoise the output, and display it at a higher res. I believe it is very similar to what photoshop calls its "Oil paint" filter. It's important to not lump in "design" choices in with what might or might not be AI. saying "this character's boots are too thin, an artist would never do this" is wrong in my opinion. There's a multitude of sensibilities in art and i can think of plenty of artists who would to that. A less experienced artists would even draw one leg "way thinner" than the other. So with that in mind a more valuable advice is "How likely is it that an artist who can create an image of THIS level, would make THIS basic mistake?". It's the discrepancy there that is the tell, not the isolated incident. The level of detail observation also falls here, as it is often an artists choice to have varying levels of detail to draw attention to an area, or it is a beginner artist's mistake to add way too much detail and texture effects to elements that do not need them, so it's not a reliable sign. Instead what detail is there, and if it's a common AI noise pattern, is more valuable. Finally, to add some extra ones: As you mentioned, repeated patterns are an issue for AI. You focused on the basket, but the diamond shapes on her armor is a repeating pattern. Fingers are repeating patterns, pipes, cables, street pavements, wheels, and many other elements. If you inspect any of these on an untouched AI image they will either be completely melting into each other, or at least very wobbly. Straight lines are rarely ever straight. Window panes, buildings, things that even the most basic artist would try to straighten with a ruler, are completely wobbly or not connecting. One good example here in the video is the handle of the sword not matching the blade's center, which is a mistake a beginner artist might make, but not the Moebius-level artist the style suggestes, lol.Similarly, circles are never actually circular. Numbers and Letters are never readable. Unless the image has been retouched, AI models never spit out actually readable text, just noise. So that's another easy tell. And last one: Is it an elf girl with her boobs out, but rendered in a painterly western style? Probably AI! xD Cheers!
Thank you! Something that bothers me a lot is how peope talk about AI doing what a lot of hobbyist artists would simply consider a shortcut. Asymmetry? Nonsense patterns? Hands? Odd small anatomy details? Ah yes, all things artists famously don't struggle with. There's a reason it's so prevalent with AI, after all. Some people may have very high skill level, but still do art for fun, so they might not bother with tiny details that aren't the focus of the piece, and just want your eyes to glide over that part. Some artists might focus on the things they personally find fun, especially at the beginning, so you might have an incongruent coloring and anatony, for example, because the artist is more interested in rendering than learning proportions, and yes, it might look goofy sometimes! The problem is that now you have people who might not know theory (like the idea of just "suggesting detail") hyperfocusing on EVERY part and "calling it out" as a gotcha. They might be right sometimes, but it still sits wrong with me.
Yeah I think a little scripting may have helped here, I kinda just rambled off the cuff and could really have used some more specific language. I'm kinda surprised people are finding the video lmfao, I've worked way harder on things that have gotten waaay less attention. All of your criticism is sound and I appreciate the time you put into this. I will however defend my critique of the skinny boot. It was way too skinny!
@@veno_net It is skinny! It's more that ai, beginner artist or personal style could all make it happen. Anyway, props for putting yourself out there for the youtube algorithm to randomly pick up. Keep up the good fight ✌
@@veno_net I would still say it was a good catch with details leading to nowhere, for example on that sci-fi armored guy. It is true that unskilled or a hobby artist wouldn't give much attention to the details they aren't interested in, however, even if just being unpolished or obscured, those details would still make sense. Most of the time, when you go through an effort to indicate some details, you do that with a reason (think of unskilled artist adding Celtic knots to armor for example, even if they are bad they would make sense in where they appear, or woven symbols in fabric, they may be just indicated, but you as a viewer would still be able to read why they are there). AI doesn't do that because it lacks context behind such kind of details, it just knows they are there sometimes. Also, textures get really weird if you look them in detail, I'm not sure how to explain, but they don't really look like a digital textured brushed ones do, more like artifacts after upscaling.
My main takeaway is that AI art makes mistakes that beginner artists would make (EDIT: should say "AI art makes mistakes even beginner artists wouldn't") but the rendering is that of an experienced artist
what if I’m really good at rendering, but my anatomy/other aspects of my art are not that great? I tend to have this problem a lot myself, and I’m worried my art looks like AI art.
Might actually help you appreciate art more. I think we usually don’t spend enough time looking over peoples art and why they put things where they are or why they used this or that color or type of style.
Welcome to the wonderful world of art. This is how people become artists. They start to notice all the little decisions, then they fall in love with making those choices themselves.
@@fallenmango8420 I have developed a reflexive version of this, but with Lego pieces. I will occasionally just start seeing Lego pieces in stuff as if I'd be trying to build a model of it.m I have not played with Lego much (not enough at least) for the better part of a decade, but it still happens every now and then.
@@fallenmango8420 this is exactly why i love art. I can think "hm. I wanna see x thing in a drawing. Im gonna do it!" Being able to make desicions, leave little details that most people will miss, do silly things that make no sense out of context like shitpost art, and then follow up with a very thoughout piece... I love that so much. AI art can NEVER understand this. They just do a whole lot of nothing. No emotion, no thoughts, nothing
I find it hilarious that out of all the material that an AI-made elf girl could gather and collect in an assumingly magical forest, she chose dead leaves. Dead fucking leaves.
the melt is a dead giveaway. even if you get lazy the stuff you draw prolly isn't gonna be melted, it'll be messy and ehhh but not... melted. because blurring and blending like that is usually more effort.
also- if you see a detail, ask yourself why someone would logically do that. the giveaways are usually melted and glitchy eyes that aren't symmetrical but not in a way a human would do but in a "this was GENERATED by a computer" way. jewelry and folds on clothes tell a lot. humans usually follow some logic. texture is a giveaway, why would a human do that? would they, would they do that at that level of skill? also lighting- if it follows no sense at all. humans and even animals understand composition and balance. ai doesn't. consistency too.
and when things overlap! usually when you draw for example a hair over an eye you draw the eye fully and match it with the other one and then add the hair and the face below isn't distorted. in ai it gets melted sometimes
Nice video! I like the term "ai spew" for those weird pockets of detailed mishmash. A few other telltale signs I've noticed: buttons and buckles that look like they've "melted", hair strands merging into straps and jewelry, inconsistent pupils, and the art is often really SHINY with blown out highlights.
@@kenicity but everyone knows Jotaro's design is from the 80s/90s long before ai. but thats where the other tale tell signs of ai art comes in if theres AI art of Jotaro. I notice many ai art of copyrighted characters tend to be inaccurate designs despite the level of perfection seen in the ai generated image
The shininess is probably the easiest tell in 90% of cases, seemingly especially when it comes to NSFW art. Another thing I've noticed with styles with notable outlines is that the outline will have much of the same issues as what you mentioned with hair, where it merges into things it isnt at all supposed to be attached to. I'm a hobby game dev so I try to keep an eye on new projects being uploaded on itch io, and nowadays the thumbnails on there tend to be infested with good examples of clearly generated images. And the "games" behind them often just boil down to whats essentially just a gallery of generated NSFW images of popular characters, either in the form of a Renpy visual novel or just a set of sliding puzzles. Edit: oh also, eyes, these algorithms seem to really dislike eyes, especially if a character is wearing glasses as well.
My art: - jumbles of nonsense in the background - poor understanding of cloth - inconsistent styles in the same painting - nonsensical hands …guys I think I’m a robot
at first glance, does your art look like it was done by someone very skilled, and that it would take a very long time to make? The main point of the video isn't to shit on those styles or choices when they are made by artists, it's to point out a larger problem with AI art (lack of intention with what might appear to be high quality art).
Because AI is not capable of replicating human observation or creativity, it doesn't understand how objects would work in reality. It doesn't understand art theory.
@@jingbot1071what pressure? companies that want properly done work will hire an artist, companies that want a cheap copycat will hire an algorithm. The only thing being truly f*cked here are indipendent artists getting less commissions
you’re teaching me how to replicate the weird lumpy organic look. thank you. i want that in my own art sometimes. ai art looks like it grew out of a petri dish there’s something so unnerving and bacterial about it. if i can learn to harness that texture in my own artwork id be unstoppable
I’d love to see that replicated by a human artist! because a human artist intentionally making their art look that way, is wayyyy better than AI unintentionally getting that kind of look from the sheer amount of pixel distortion from merging hundreds of stolen images together.
I think a big part of the problem is seeing art as a purely utilitarian thing where the results are the only thing that matters. A lot of Ai enthusiasts only seem to care about the immediate gratification of seeing "Sexy Emma Watson dressed as steampunk Ironman ultra realistic 4K" before they move to generate their next super original idea. For me the saddest part is the people that devoid themselves of the satisfaction of learning and appreciating the process and by consequence loose respect for the craft. It´s very difficult for them to grasp that we don't revere artists for making "pretty images" but for finding and developing a voice to make their soul speak in a visual language.
art has always been to me almost more about appreciating the vision, the creativity, the skills etc. of the artist, so seeing people just generating images and finding them so amazing just tells me they don't have any actual appreciation for art. really I personally don't even consider it "art" if it wasn't made by a human. it's just... an image. that's all. and as a result i can't find myself appreciating the images in the slightest outside of "yeah that's pretty, i guess"
I love your entire comment, and the last couple of sentences in particular. I said this somewhere else but I hope mixed media becomes more prevalent. AI cannot harvest cloth, dried herbs and materials from a trip it took and make something meaningful out of it. Yet that's the art that moved me most
I totally agree with you, well said. Art is about connecting to your soul/the source/universe and creating what you're inspired to create. The process is often more important than the finished piece for the artist. People who appreciate art and craft love watching it being created. Ai doesn't create art, it needs to be separate from true art. Maybe we can call it human made art or soul art so that people understand that it has a soul connection and wasn't created by a robot.
This is something I think it's really hard for non-artists to understand. I can't draw for crap, and I really don't have much interest in it, so AI art becomes a very easy way to jump right into the final product part. As a singer though, I can imagine what artists must think about this stuff because of how I feel about AI music. It's obvious to a creator that the creation process is where the joy is, not so to people who are more strictly consumers.
While its true that "ai spew" is a pretty reliable giveaway, i think its worth mentioning that some artists that prefer to work sketchy tend to end up with similiar "spews" at glance. I myself tend to create nonsensical designs out of the top of my head in places where i think no attention is needed and space should be filled, just to give a general visual sense of what i have in mind to the viewer. But if you notice common patterns in ai spews then yeah, it becomes obvious, otherwise it might as well be artists sketch work
The difference between you and ai is that the ai is not "working sketchy." Most if not all of the pieces showcased would have taken a human weeks or months, and simply not noticing or caring that a hand is holding a piece of nothing or that an area is filled with pipes that don't go anywhere ceases to be a reasonable explanation. I found the first piece he showed the most believable, because at a glance it looked like someone throwing flats and subtle shading onto an otherwise very sketchy piece. The fact that most of the nothing is supposed to be ruins also helps. I feel like the sword hilt and the boot are the only real givaways (the best givaway is the bit of sword hilt on his arm).
i wouldn't say so, it's obvious when it's a sketch or mistake or laziness or human error or action- what humans do is usually intentional. varying levels of intentional but definitely still lucid unless you're on complete autopilot or not sober. what ai does is clearly unintentional and random, melted, fuzzy. complete nonsense that is easy to notice if you train your eye.
I think one of the most effective ways to recognize AI "art" is simply to expose yourself to it. Look at it enough and you wil lstart to see patterns, shapes and colors that make it almost immideatly obvious you are dealing with a generated image. Be it realistic, stylized ect theres always indicators in all of them that will stick out to you the more you see them.
Even better. Try it yourself. My brother who's obsessed with rats was making a ton of ai rat images on Midjourny. In ones where he tried making crowds of rats you'll see them get copied and pasted and melt together into rat soup. He let me try my own promts and as a curious artist. I wanted to see mythical creatures that rarely ever get drawn. I'll describe them the best I can and the Ai wouldn't even get close to it and would even make random stuff that has nothing to do with the prompt at all. After that I felt pretty secured that the stuff I'm most passionate about drawing can't be replicated by ai.
AI art is kind of like how you imagine a very complicated art piece in your head when you just imagine it at a first glance it looks so cool and amazing but when you try to flesh out the details of how would it look like exactly, everything just breaks
I think the main takeaway here and why so many people fail to recognize AI "art" is the fact that most people don't know how to truly appreciate art. They will see a piece of AI art online, or any art for that matter, and look at it for maybe 5 seconds, think "wow that's so realistic! I could never do that!" and move on. Few people actually stop to take in the details and appreciate the time, skill, and effort it takes to create a piece. I wish more people would treat art in online spaces the way they do in an art gallery - not rushing past every painting on the wall, but really stopping and staring and noticing.
Is that necessarily bad tho, for most ppl art is a decoration and doesn't have a meaning apart of aesthetics. Like if you look at a curtain can you see a deeper idea in the floral pattern, no. Not everyone cares about art and it's not a wrong thing to do.
@@AGM_V2I think that is kind of wrong, actually. consumption for consumption's sake, whether it be AI art or floral patterns on curtains or endlessly scrolling through tiktok is a bad habit to be in, albeit to different degrees. i do think it's been normalized, but in the same way that alcohol and cigarettes and fast fashion have. we've gotten very used to this shallow, aesthetic relationship with art and I think that's a lot of the reason our media, our clothes, and our furniture is cheap garbage (doesn't help that the material stuff also produces lots of waste!). it doesn't have any real purpose besides existing for the sake of it. not to be seen, felt, or understood, just noise.
this is totally right on the mark. I find myself missing the details that would clue to a work being AI art, and it's always because I just glanced rather than actually taking in the work
They're not scared. They're lazy. They don't want to put in the effort, and they're satisfied with the fantasy that AI fulfills for them. My opinion as an artist: what's going to end up happening is that the bar will be raised. That "caliber" of AI art is going to become the starting line and real artists are going to aim to surpass that bar. Artists will focus more on illustrations with a story--with context, theme, and narrative--and work even more to build a personal style with a solid reputation. They will also focus on designs with authentic organic patterns -- the kind that is very hard for AI to interpret. Personally, I'm not a doomer about it. The real advantage AI has is speed and output. But until they create an AI with the same capability as the human brain, a real artist is always going to be more skilled.
Your standards are valid but close minded, because others can express themselves differently you might want to take something away from them? Would you agree that there are monster artist out there that set a standard so high it's almost untouchable? Are they "Doomers"? Did they steal all Art from you? No, Art is the size of of the universe, we have many many worlds to explore, it shouldn't be about conquering, it should be about exploring. [Why do I sense you're hinting at financial gain instead of the love of art]
@@Hahshdhbcbcyoutube And how do you think any actual artist starts learning ? They would also look at others art and take inspiration...hell, sometimes imitate things they may like as well in their own art. Everything anyone creates is inspired by what they have seen in like which guess what ? Also includes art of other artists.
@@amadeusnagamine9056 I'm so tired of this "but artist use inspiration just like ai do! Bla bla bla " Well here's the problem, ai can make a perfect replica and humans just simply can't. No begginer takes, for example samdoesarts work and make a perfect copy of it. Please watch more video about this topic instead of bringing the same thing to the table
@@Hahshdhbcbcyoutube *"ai can make a perfect replica and humans just simply can't"* As of your knowledge, I'll grant you that it maybe the case as of yet, but not in the future. Humans are amazing so don't hastily steal away what we might know yet. *"No begginer takes, for example samdoesarts work and make a perfect copy of it"* Agreed, give them time and would you agree they could? So time is only an issue here? So my understanding what you may be trying to express is: The time it takes for A.I vs Human to create/learn how to do a piece of Art.
My advise if it's your concern, is everyone and everything that can learn, learn at different speeds, even now A.I. is "learning" and improving and adapting. My point above is, what is your goal my friend? Is it to conquer or is it to explore. (One want's it now, and the other is timeless.) So? To conquer or is it to explore? (or something else?)
I’m not a professional artist, but I draw for fun. When I see AI generated art it gives me an uncanny valley feeling. At first glance it looks good but SOMETHING always feels….off. As someone that does this for fun, I like the idea of AI helping me get the repetitive work done, or helping me do the time consuming stuff.
For the first one, it wasn’t the “spew” of the AI that tipped me off immediately, since that particular piece was clearly generated using Mobius as a prompt- and his style is known for that sort of blending of organic shapes with technology. What jumped out to me before you even zoomed in was the weird perspective and angling of the sword. The handle is crooked, like you pointed out but also looking at where the tip rests on the ground, it’s just not quite right for how it should sit in a physical space.
"AI SPEW" is such a great way to describe that whole detail creating nothing tendency that AI images have. If you can look at some bit of detail and go "yeah, but what is it supposed to be?" that's usually a good indication of AI art. There's lots of other indications that I come across that are obviously AI, but a lot of those feel barely worth mentioning because those will get better in time, but the things I think will NOT get in better in time (or at least not as quickly) is something that mainly accomplished real artists will pick up on. It's when the image is rendered to a high degree, but also wrong in some fundamental ways, and any professioonal level artist will instantly see that any human that could paint or render to that degree would not still be making those kinds of fundamental errors in light sources matching, or perspective, or anatomy, or a thousand other tells. "AI artists" who claim that "people can't tell the difference" just have acase of not knowing what they don't know, because sure THEY can't tell the difference between human art and AI images, but certainly accomplished artists who have seen enough AI art can see it pretty clearly.
"any human that could paint or render to that degree would not still be making those kinds of fundamental errors" Yeah, I think this is the biggest tell. Simply the distribution of errors is nothing like the distribution of errors in a human drawing. Sure, all those mistakes that AI makes could've been made by a human, but the overall combination of "awesome" and "awesomly stupid" in a single illustration is different when done by AI vs when done by huan.
AI can now replicate text convincingly well under the proper conditions. It now understands how hands and feet are supposed to look. It can tell the difference between hair and the environment. But it can never replace human creativity. Poses are something it seems to still struggle with, especially very dynamic or action oriented poses.
@@ghostoflazlomost people are bad at using prompts and don't use photoshop to touch up flaws in their generations. Drawing something gives you full control over your final result instead of being forced to correct things and generate stuff constantly. AI will always fuck up something in the process
3:29 One little thing I want to add to this notion about how a real artist wouldn't make that kind of error- It should be specified that an artist of that particular skill level wouldn't make that kind of error! Generally, you need to think about how the presented errors compare to what skill level the piece would represent had it been made by a real human. If the art and design appear to be of professional quality and skill but has errors like this, then yeah, something's definitely fishy there. But if the art looks to be made by someone who not nearly as skilled, then it would more likely just be a matter of an error made by someone who's simply still learning. Of course, it's still very sad that we even have to scrutinize any good-looking art these days so we don't accidentally support lazy AI users who think messing with prompts takes the same kind of effort and skill as making genuine art.
And I hate how a lot of people call themselves AI artists. There is no such thing as AI artist, what is artist about it, you write a prompt and you're the next Van Gogh?!
@@Renikee Yeah there is literally zero difference between an "AI artist" and someone just googling "cool space drawing" and finding their favourite image. They have put in the exact same amount of creative input, it's just one extra level separated from the original creator. The problem is the exact same type of person who thinks AI images are art that they created, also have never experienced the drive to create art in their lives so don't actually even know what it feels like. They describe themselves as artists, but they actually don't actually understand what that means.
I hope there’s more regulation out around what kinds of images can be fed into ai image generators in the future. The biggest fear for me isn’t so much that ai art can look like human art, but rather that most of these image generating softwares like Dall E and Midjourny use an unethical data scrubbing system that takes data from artists online without permission or compensation. It’s unbelievably exploitative.
I hope the current lawsuit against Microsoft and OpenAI will have a domino effect on all of these other AI models. Maybe the datasets used to train AI should be limited to what is available in the public domain/creative commons or at least purchasable from stock media sources.
@@enviritas9498 it would just make the AI sound like its from 1850s. It would really look good. Would you want to stop progress just for that? What about the science?
There was a post of conversation with the midjourney creators and they had a whole list of artist they stole from while encouraging additional names to the list. Meanwhile theyre claiming to media they dont know who they stole from.
@@TheManinBlack9054 What kind of progress does AI image generation help with? What science? Art is a creative outlet, it's inherently human. Take away the human and it becomes nothing more than a mesh of patterns and color.
One of the things I notice the most is that because image generation models make everything in a shotgun blast of everything all at once, it has a similar effect to a person creating it with immediate amnesia of why it started drawing any particular line. They end up continuing and sharing lines between different component pieces. So you end up with these weird lines that start definitely as part of one element, that just all of a sudden become definitely part of another. The model has no idea that these are two elements, it just knows "lines continue" and "line goes here" - how that interacts with the rest of the composition doesn't weigh in to its decision making. From the model's perspective, it has correctly fulfilled the request of having each pixel or group of pixels make contextual sense with the ones next to it. You can see it happening with the 1st image's right boot. There's a line that starts definitely as a fold in the leather, that suddenly becomes definitely a detail of the ground between his feet. The 2nd elf woman's basket has lines that start definitely as the rolled edge, that suddenly become definitely strands of twine.
I saw a video where an expert was explaining why a faked photo of a supposed armoury generated things such as rifles with barrels coming out of both ends of s stock, or multiple magazines sticking out of it: the software knew a barrel and magazine was supposed to come out of a rifle's stock, but "forgot" that it had them already coming out one side and so cheerfully kept producing them.
I love this. As an illustrator myself, I always found it relatively easy to tell when art is AI generated, almost like second nature, but I know that for a lot of people that’s not the case. I think this video did an amazing job explaining the logic of it all, and put into words very clearly an instinct a lot of artists might already have when analyzing if an art is AI. I hope this video blows up so more people can also learn how to differentiate AI from real, genuine human-made art.
Yeah initially people were amazed by ai art. Now it's like it's hard to imagine ai art being used anywhere legitimately because everyone knows how to notice these "suggestions of something". And once you notice just one of these in a setting, it's a turn-off.
Quite a few companies have tried... For example, I still have no idea how Wacom of all companies, an ART based one at that! Posted Ai images and used the excuse "We didn't notice it was ai" When the detached dragon tails and fuked up teeth were sooo obvious. Lol
Sadly i'm not that sure about this.. I see lots of book covers in retail that are obviously AI Generated yet it doesn't seem to bother most of people (and certainly not publishers c: ). Same for birthday cards or whatnot. Lately Palworld a game that is really popular on steam (sigh) is a mishmash of asset flipping and IA Generated spew. Lots of people don't have the visual baggage to differenciate the AI spew from real art. I wish people would notice and be turned off but i'm not really convinced it will work that way sadly :/
There's AI art winning art contests mixed in with traditional art. Also this is AI's early years while humans have had millions of years to brush up their game
@@karrawr9538AI’s artistic limit is about wherever the current limit of humanity’s skill is wherever the time period it’s in as it’s only capability is to mix, mash, and combine already existing styles into an image with zero personality. The only real thing it can do as of currently is to refine rough edges and integrate prompting more accurately, other than that, it ain’t going far.
I think AI art is just like fast food , it only appeals to people who wants the dopamine rush to look at something for a small amount of time which is one of the reasons why AI art is so popular in social media sites like instagram . Only few people can enjoy art and spend time collecting and provide valubale suggestions for them .
Nah its more like matrix nutrition paste imitates glucose and protein but has synthetic snot taste but they trick themselves into thinking its mcdonald food then its not even that
i see your point, but unlike art, food is a necessity and some paychecks don't allow for good quality food so that doesn't really equate (lighthearted). can't say this is a much better analogy but i'd compare it to scrolling on whatever youtube's version of an fyp is vs watching long-form videos. you'd rather get handed whatever the algorithm thinks you want as quickly as possible regardless of quality than search for a creator you like or a topic you actually want to engage in.
AI will have that "too good to be real" look from a far but up close it has this unclear, foggy, and sometimes weird look to it that's super common among AI images (especially when it comes to humans). I am really curious though how exactly AI produces art like this.
it’s crazy how whenever i try to search google for pictures of any artistic sort, 90% of it is AI generated. it’s really getting on my nerves, and as much as i was thrilled about the usefulness and capability of AI at first, i now started to hate it more and more.
the good side is that will make AI worse, because it steal photos from internet, and most of these photos being AI... well, i think you can understand what will happen. (sorry for the bad english)
Im an artist as well and I have tons of issues with AI art, some of which youve mentioned in the video, but one of the biggest and most saddening things i hate about AI art is it has forever changed how we can view and enjoy art. We can never just see a beautiful artwork and immediately be in awe of the effort and skills of the creator. I remember i used to love just going on Pinterest because I loved seeing tons of amazing, incredible art and adding them to my boards which would continually grow as i continue to be suggested more and more beautiful artworks. Now Pinterest is just filled with AI garbage and is just a site that reminds me of the exploitative technology. It's like visiting your favorite beach and suddenly realising the water is full of plastic and trash...
Totally agree. I do use those AI arts from pinterest as reference and use my own style and stuff i use and make my own. Im stealing from the stealers >:)
Thank you so much. I'm not a digital painter and not much of an artist anyway, but its nice to know how you recognise it. To be honest, in many cases I would've been fooled... I would've made excuses for the artist, saying "he didnt know better" or "he must've thought something, I'm just not seeing it" ... I still find it hard but this certainly is a start.
On the Sci-Fi/Fantasy art- I seen a guy popp off, a guy who from his writing seems to be quie a good writter, but they been using AI for illustrations and you could notice the "badass art about nothing" vibe you were talking about. Plus most of the "art" output looked like if it was mildly trained on Fallout concept arts. I was mostly disapointed in the guy, could have scribbled things themselves or get someone to do it for them- and the "I ain't have the time" is a fkin insulting cop-out
Why is it a cop-out? I don't see a good way to achieve his illustrations. His own scribbles would prob. have looked horrrible and put people off. Finding someone to do it for you for free is quite rare, you would likely have to pay for the commission. Apart from the fact that the model was probably trained on images their artist have not given consent to (which is immoral), his approach seems pragmatic from another writer's perspective.
One of (edit: The only) beneficial things of AI people posting a zillion images a day is that it's kind of self destructive to generative AI. Since theres sadly no real reliable way for an AI to look at a billions of images and categorize them as AI or not, when they go to scrape and steal images for training data they'll inevitably end up with tons of AI crap in there. (And much of it not tagged as AI by the poster because many AI people dont want to open and honest a computer made it) Meaning errors or bias that propagate in the output might be magnified because they're now present in the input. Not sure how these AI companies will deal with that because it's a problem for them that'll only get worse with each passing day.
i draw in a quite realistic style with some elements of abstraction (a lot of randomized pencils without necessarily drawing a sketch or specifying what it is) and i get often accused of being AI, it's really annoying. The AI copied me, not the other way around.
The immense output is a good way to spot an AI prompter, as well as checking the art style. A person's art style will evolve naturally over time, but if several pieces posted within days of each other each look radically different, that's a red flag. (These are the same tips commonly used to catch art thieves, which, let's be real, AI "artists" essentially are. Some things never change lol)
I feel like the flood of IA generation we're getting is going to loose it's worth quite fast in the long run. Those people generating art assume that the human eye cannot be trained to recognize what's human from what is not. Fun fact: it is. Just like with any new concept we discover, it takes time for us to "adapt and get used to it". But after you understand the general parterns and repetitions of what an IA technically capable of, (by also deepening your culture and knowledge allowing you to recognize which human artist's work got fed into the machine) then it becomes insanely obvious and jumps to your eyes the second you check those "art accounts" :)
It's also gonna lose it's worth in the long run due to accidentalyl feeding AI generated images back into the training data, which reduces the output quality
@@EkattPalladium The rest of your post is in English, and you're commenting on an English video that uses the English acronym. Don't pretend to be ignorant.
Another thing that generally stands out to me in AI art is just the choice of color, usually. Generally it looks very uniform and uses the same sort of dull, desaturated colors. It may be purely anecdotal on my part, but I always take a look at the color. Light sources, too. To preface this: I am not an artist, the most I've done art wise is just the funny channel avatar and my profile banner. Not particularly great work (especially since the cropping on channel banners made it hard to include the full piece that should be there). But it's still possible to notice these details if you have some clue as to what you're looking for.
No I am an artist and I pay attention to this stuff and I can’t tell you because I’m deep in this I can tell you artist that I study by name I have a whole Pinterest dedicated to art and things I want to buy and other things I’m interested in and ai can create very vibrant pieces of art the best way to tell something is ai is the eyes of the characters are very fixated and the faces
I really enjoy what you call the “AI spew”, when you have a hint of something that turns out to be an illusion, this is what I strive for in my traditional painting.
Commenting to boost the algorithm, this video was very good !! Telling apart AI from real art is super important, definitely sharing this with people to spread the info. Sending all my hugs and good vibes to you !! Also, ive seen some people defending AI generated images as "art" by saying its "making art accessible" as if theres something stopping them from picking up a pencil and a piece of paper ತ_ತ they act as if art and creating is like a coveted craft only certain "talented" people can do and its really just sad, because it shows how little will to practice and actually care about what they do they have. Anyway, just a little rant from me lololol :)
I fear that a lot of up and coming artist, myself included, face the hardships of dealing with AI in the industry/ art space. Already don't like that fact that discerning what is AI and what is actual art is a problem especially for someone who is not well known. Even worse when you see a professional concept artist start to use midjourney, then it comes into question who is mixing it in their work or not. It's such a mess. Great video btw and hope you continue your art journey!
I mean a lot of artists have no imagination, go only literally and art site sort by new with the AI filter on and 95 percent of what you see is bad art and uncreative compositions. People like to compare AI art to the top like 5 percent of artists and not the sea of absolute schlock that people actually tend to put out.
@@f556784q3I think this is cope. Even if it may not be some Picaso or something, at least those ppl actually take the time do the thing themselves. Every artist starts somewhere and shows some drive and passion. "Well if you think thats bad this is worse! They lack imagination so that means I do the same thing but with less effort by having something else do it for me and only works because of said artist!" . It's playing the mediocrity game.
True. Ai "art" can generate some really amazing weird stuff if you let it. But most of the people using it generate the same generic images of hot fantasy girls
people think ai is a shortcut to cool art. they get a rush when they can instantly manufacture a flashy image. but that's all they get. a rush. no greater understanding of their subject or how they relate to it.
The way i could tell whether its ai or not is when there are ugly random glitch colors and and the ears with the earings and the weird clumbs in hair (which is usually FILLED with hair strands) i dont really have a problem with ai when its not used with ill intentions like a paying customer for a comission. Some are great for giving ideas
I feel really sorry for artists. I once dreamed of being an artist, of learning how to digital paint beautiful fantasy art like frazetta, brom and guys like that, those D&D artworks stuff, also scify... all the beautiful things they were doing out of their uncanny talent they built over decades of studying. And now their work have just become replicable by ai in just a feel prompts and a couple of minutes. I feel really sorry for that. Had I chosen this path for my life, had I followed my dream, I would be devastated now.
You know in a way, i think AI might make the art community stronger in the long run. Right now we are in the thick of it, the chaos days of early AI havoc, but i think a stronger vision and respect regarding what art is may come out of it. This may take time though, years. In a way, there is huge engagement in art right now, all these people doing AI and people talking about it. They will not keep doing AI. But those who stay after the fad dies out, realizing they haven't built up anything in their years using AI, may start doing actual art.
ultimately it depends on the people. the creators of midjourney have been caught with a discord screenshot proving they're plotting to launder money and dehumanizing artists (if you need the link i can't post it because youtube doesnt allow links i think) but im sure you can find it somewhere. im sure ai art will help a lot but right now yeah it's a cesspit
@@lemonadesaccounttm8721 wow that's messed up. But nah i don't need to see it. It's very much about the people yes. There will always be bad players. But the good ones will prevail!
@@itsgonnabeanaurfrommeI wouldn’t call it entirely stupid. Google became such a phenomenon that it became a verb. Amazon also defined online shipping sites. They are some of the biggest, but partially because they were some of the first. With the world we live in, a company can absolutely be representative of a technology, especially in the short term, where they are one of the only few ones widely supplying it. Besides, the replier had a point. Midjourney is one of many AI ‘art’ programs whose users consider art nothing but a product, and right now, it is a cesspit. And I agree with both the replier and the original commenter, that it might have future uses and be a net positive. But for now, it’s causing harm, and Midjourney is just one example of that.
From my observations, AI art usually struggles with consistent lighting. Primary light sources are affected by bias from the input data, the light source can be behind the character\object, but the front will still be lit because artists prefer to position their light in certain ways. Shadows are guessed\approximate and not consciously projected. And bounced light usually incorrectly represents how the light would interact with the surfaces in the scene. There are some ways to improve it, for example using a simple 3d scene from Blender as a rough input but the result would still be guessed based on the collected patterns and biases because that's how neural networks process information.
“It loves art about nothing” is a perfect description of AI-generated stuff 👌 That’s exactly the feeling I get when I see it, and it immediately signals to me that no real human was behind it (except for the artists that the AI was trained on).
Thanks for posting this. It needed to be said, and it will need to be said many times over. I'm a non-pro digital artist who spent years building up my own style, that sadly is very wispy/dreamlike, and therefore quite easy to be drowned out by AI. I got pretty despondent about it at first, but I've since turned it around, gone back to learning how to tighten my compositions, but also how to use oil and other physical media to augment my work. When I start posting again I will include images showing my rough work and progress to prove that it's my work. I totally agree with you that AI cheapens art, and that people who use it don't seem to realize that art is as much the process of creation as it is the result. But I think it will cause people who truly love art, artist and fans alike, to appreciate that work more, and to value and support people who are doing it for real.
Another thing I’ve noticed, usually with AI pictures that have multiple characters, is that the lighting and color between each character and the background would just be slightly but noticeably off, like a bad green screen where the lighting on one thing is very different than another.
I draw my more detailed scenes like the first ai lol. I work off a lot of random shapes to imply things and just hope that people don't look too hard into it. It's one of the reasons I generally don't draw whole scenes. I also am just awful at planning so that's probably another reason I throw random stuff in lol
One of the saddest things about AI for me is I can no longer just enjoy art, when scrolling through pinterest looking at cool art I'll always have to be suspicious of everything Before all this I could just scroll and be joyful at others creations, now I use pinterest and just feel empty
My biggest giveaway is the fact AI art looks too 3D and polished. Another big giveaway for me is the mistakes AI makes sometimes. Eg- 6 fingers, 3 nostrils, jumbled up text like "frnczon 2"
I really love your take on this and you put it very eloquently when you said that AI "art" is made by people afraid of trying, afraid of failing, afraid of being engaged in the art process. It is a sad day we live in as artists and I worry for the future of artistic integrity. I think traditional media will have more value in the future as digital art becomes more and more saturated with ai generation. The frustrating and sad thing is that we are increasingly seeing that the people who care about art are artists. Which is to say that the majority of the populace just care about the picture they look at, not the way it was made. They want to see something pretty (hence all the subjects of ai art being pretty women/men in popular themes like cyberpunk) but don't care how it was made.
Seeing this video pop up on my youtube feed was nice, because I see so much AI art on youtube and other places, it makes me mad. Thanks for helping me identify weather it's real or not. I respect the passion and time that goes into making real art.
Good indicators are also "AI scribble" where a bunch of letters got layered over another to create the illusion of text, additional comes "AI grain abstinence" which is basically just the absence of luminecant noise and color noise, when you turn down these two facors to a max in Photoshop your handmade picture will look more like made by AI, the very smooth "non-texture" is rather characteristic for AI imo
You're so good at pacing and speaking in an engaging way that when I started this video I was like aw man 18 minutes I'm not gonna sit through that let me just add it to my watch later but I started to watch a bit just to get a taste and suddenly the video was over and it felt like no time at all.
I actuallt tried generating some ai art(not posting of course) just to see if I could notice the details, and the mismatched style is actually such a large giveaway, it was really blocky and suddenly got photorealistic for the skin, this is super helpful
I'd be very keen on seeing you do a blind Spot-the-AI video. Say 7 images, chosen by another artist, and you examine them to pick the fakes and see how well you do.
I once even made an acrylic painting of pokemon Tauros inspired by Van Gogh painting called Lying Cow and I wrote "to prove it's not AI I'm add the picture of my work in progress painting". I think it's important to add work w.i.p. for some art to show people that we not make an AI art.
a lot of ai artist also chose topics that are very common for beginers elf lady is legit the firs thing i tryed to draw when i was 9 and decided to learn and i have multiple drawing from elementry school with the exact same posing as the guy with the sword it's just a little funny to me how all beginers seem to have the same ideas of what you shoud make
This video just put into words a lot of thoughts and feelings that I've had for AI "art" for a long time. LIke the blending of styles, and the inconsistency of photorealism and surrealism in one picture is something humans would rarely if ever do.
Someone said that “think as an artist”, see any detail and ask “if I’m an artist, why did i put it there, why did I draw this detail” after that it is easy to see ai art
The example at 7:09 is just wild to me. I feel like a crazy person telling people just how CLOSE some of these outputs are to the original art and/or each other, NONE of these GANs are producing 'original' work, its pieces and bits closer or far away from a small pool of originals that it grabs within a processing loop. There's really NOTHING to be 'impressed" about here other than it's 'interesting' how this method to processing iterations of pieces between bits of noise and grain are how you get a derivative result, and the speed in which it's processed. IE: NO ONE is impressed by taking a photo and putting photoshop's "Oil painting" filter on it, it looks like shit. But essentially, that's what GANs do, just in a more 'seamless' way, by setting up a 'composition' that's practically a photobash, and adding or removing 'eased in' elements from other photos either a little or a lot. Just like the photoshop filter, this is done through using the noise that is present in all images through the color data / pixels The fact people genuinely believe GAN assisted imagery is created from scratch, pixel by pixel is so ridiculous to me. Rant aside, awesome video. Doing my part in watching and commenting. Just passing through.
@@cccbbbccc5910 Rendering is the primary metric most non-artists use to judge art, but creativity is rarely considered. Think of a prolific artist or studio with a signature style who spawns dozens or hundreds of imitators. We can easily spot the influences of these artists and recognize them as imitations. Most of these imitators are unremarkable and don't progress the industry. The element that makes an artist stand out among a sea of imitators is what AI will never have: creativity.
When I draw or watch artists, a lot of them do like beauty, but, there is always method, usually. Like, the folds of the hair and clothes usually have a pattern to them, and a method to them. Like, Shinichi Sakamoto loves to draw bishoenen and biseinen, but there is a lot of method to his characters.
I've always been a huge fan of weird or horror type of art. Unnatural stuff that most people would never think of. I love some of the tnings ai spits out sometimes. Like the weird optical illusions. It looks right but something is off. Makes you feel uneasy. Legs turning into arms. Small details that sometimes are so creative in a way that a human would never think of. Especially if youre making pornographic stuff but then suddenly one of the pictures is just creepy and horrific. Always been into fantasy art and horror art. Disturbing stuff reallytaps into your emotional responses. Ai does this so well. Its mind blowing.
Keep in mind, There are alot of false positives. There is currently no relaible way to determine if art is AI, and beginner artists can make alot of these mistakes. (I know I do) Drawing pretty woman, Having bad folds in clothing, and having a boot that is too small, having an inconsistent style, are all things that I would do lmao, and I have been drawing for ten years. Also, AI art in skilled AI creators (those that use control net, photo editing, LORA, ect) you wont really be able to tell.
im definitely guilty of the ‘pretty women’ thing too, but ai doesn’t it in a very specific way, usually a painterly anime-esque style with fantasy elements
@@elvenbugs one of my favorite artists, wlop (the creator of the knight shielding princess meme) does it in exactly the AI art style. Kooleen does this too.
You perfectly explained why AI generated images are so uncanny, the fact that it generates organic looking globs of nothing that gets more frustrating to try to figure out the longer you look at it. Because it looks so familiar but at the same time makes no physical sense. Even in the setting of a fictional world the background usually starts breaking the farther back it goes. It’s incredible technology don’t get me wrong and we’ve already been using a form of it for years with photoshop but at least there’s still work and effort and style put in. I really think that AI generated images should at least have a watermark or sign saying that it’s AI, because a lot of people are fooled easily.
Before I start and for context's sake I would like to emphasise the fact that I am a computer scientist who dealt with AI on a foundational level (not just prompts). I see all this well put critique as valid at the time of your recording. Thank you, I've subbed. IMHO it will only be a matter of time until things will get sorted out on the AI generative front, making it far harder for humans to tell the difference of what's men made and what's not. I don't share the optimistic opinion that good artists will leverage these algorithms to boost their quality, throughput and reach. Proof can be found in the content of your video itself. These tools are still way too fresh for the collective consciousness to have them absorbed and know what to do with them (and, arguably more important, what not to). Psychologically speaking, this does explain why there seams to be a a natural incentive (call it hype, if you will) where more and more people will try to cheaply capitalise on these low-effort cash-cows that are art works. Instead, I do believe that, for a time at least, human-made art volumes will go down (to what extent, I don't know). Many artists (known and unknown) will feel the burn and will probably give up the field altogether. It's not all doom and gloom, though. Art constantly needs to refresh itself all the time to stay relevant and have an appeal in the human psyche. Having no real art to feed the machine will inevitably lead to a degrade in quality of the AI's generated content. This will get noticed and affect its viewership which will again allow for humans to come and reseed the field. How that artist will look then it's anyone's guess. I do have a hunch (or rather a hope) that when it will re-surface, it will contain some sorts of nostalgic vibes woven into it. Of course, if we will manage to build machines that physically experience the real world before that tipping point will be reached, I do think this causal trajectory would have to be revisited. If this does happen, though (can't tell the odds, but they're not improbable), and machines will be able to generate content based on experiential input, we would find ourselves in a totally different philosophical realm of existence that would weight heavily in favor of artificial consciousness being achieved. The series "Altered Carbon" portraits a really cool concept of a future in which humanity avoids AI just because it knows it can't escape it's optimised purpose of existence even though, ironically, humans were the ones that brought them into existence. A likely pandora's box ending scenario if you ask me. Thank you once again. Take care!
Sorry, but I just need to say, I feel like saying AI art preys on people's apathy toward learning how the world works is a little harsh. It's pretty obvious to me that the reason AI art is so prevalent is much simpler than that: it is designed for the art sensibilities of social media. That is to say, it is easily consumable and looks passable at a glance, so most people won't investigate beyond that. You could call this a kind of apathy, yes, but only the same apathy as we feel toward most other internet "content" - that is, an apathy which is merely a symptom of the modern age and, in my opinion, does not reflect poorly on a consumer's character or their interest in the world around them. I am certain someone could be familiar with how clothing or mechanics work and still miss some of these giveaways that seem very obvious to you as a digital painter. I saw you state in another comment that this was off the cuff and could have been worded better if scripted, and that's fine. I'm not asking anything of you with this comment. I just thought it was worth saying.
But can you deny the world general population is becoming apathetic, even in real life? I would agree that the apathy is here long before AI Art, but wouldn't AI art feed into that apathy and encourage it to grow? Would AI arts inspired anyone to be an artist, in the same way a bland yet beautiful anime art style did? Is AI Art making effort seems worthless? We used to be inspired by one another artist and have a discussion about it, now what kind of discussion will we have instead? What will happen to the next generation? The one who wouldn't experience the world before AI Art? This is about iPad kids with worse fine motor muscle, who can't write well, can't focus well, can't communicate well.
@@EarthWingedDragon Congratulations, you now feel like everyone in every other industry. "What will the next generation do?" Torch a lobbyist or two if they know what's good for them.
Not sure if it’s in all cases, but AI can’t generate hands or feet properly. Sometimes there’s six fingers, sometimes four 😂.. so that’s one fast way to tell.
It's largely based on the model being used. The people who actually put effort into their generations (iterative, manual cleanup, etc) figured out the extra fingers thing ages ago, but people using Bing AI or Dall-E without knowing how to properly prompt will probably still get them.
Another thing ai does a lot, at least right now, is that it feels like there's a color filter on everything. Everything is desaturated such that it looks like someone put clear paper over it. If you take these images and manually turn up the saturation it looks way better and like that's how it was supposed to look.
hi veno. I'm an illustrator myself (leobrunogkk on Artstation) and I would like to give you some feedback on this video, and express some concerns.
First of all I think it's a very good idea and it's important to have these videos. Some of the warning signs here are great, unfortunately i think others may very much cause confusion and put both beginners and pros in the targets of people who take your advice the wrong way. This post will be long, so sorry if it gets boring, but i can't think of a better way to structure it.
The opening section about looking at the gallery is definitely the most infalible method you suggested. I would like to add to your "Subject matter" and "sheer volume" the amount of styles and how neatly chronologically layed out they are. A working artist, beginner or professional, even if they work in many styles, will generally either have a clear progression or have a scattering of styles and colors through time. An AI prompter will have neatly repeated blocks of identical color and style and subject matter in their galleries. Easy tell.
The "AI Spew" section is also definitely spot on. When it comes to intricate detail it does devolve into nonsense when the AI does it, but I would like to make a distinction in this "nonsense". A lot of artists who are not working in a professional capacity, or are posting works in progress or speedpaints will have unfinished details. Not every artist is master of mechanics, or foliage, or other types of detail, so I think it isn't responsible to say "If this detail doesn't make sense, it might be AI". Until we're at a certain level, a lot of what we draw "doesn't make sense".
Instead, there are very specific noise patterns that AI tends to recreate, and you glide through a few of them during your video. One of these patterns is the "soft-scaley" noise that can be seen in the Elf character's white undershirt. In the same frame it can be seen on leaves in the background, the edges of her leather armor, and other places. A different noise pattern that is recurring is an almost HR Giger spine-like repeating pattern that can often be seen in more "realistic" images (it's very visible in those recent MTGA promos). Another noise pattern that is in basically every AI image is the "Oil Paint" effect, which occurs from the algorithm each model uses to denoise the output, and display it at a higher res. I believe it is very similar to what photoshop calls its "Oil paint" filter.
It's important to not lump in "design" choices in with what might or might not be AI. saying "this character's boots are too thin, an artist would never do this" is wrong in my opinion. There's a multitude of sensibilities in art and i can think of plenty of artists who would to that. A less experienced artists would even draw one leg "way thinner" than the other.
So with that in mind a more valuable advice is "How likely is it that an artist who can create an image of THIS level, would make THIS basic mistake?". It's the discrepancy there that is the tell, not the isolated incident.
The level of detail observation also falls here, as it is often an artists choice to have varying levels of detail to draw attention to an area, or it is a beginner artist's mistake to add way too much detail and texture effects to elements that do not need them, so it's not a reliable sign. Instead what detail is there, and if it's a common AI noise pattern, is more valuable.
Finally, to add some extra ones: As you mentioned, repeated patterns are an issue for AI. You focused on the basket, but the diamond shapes on her armor is a repeating pattern. Fingers are repeating patterns, pipes, cables, street pavements, wheels, and many other elements. If you inspect any of these on an untouched AI image they will either be completely melting into each other, or at least very wobbly.
Straight lines are rarely ever straight. Window panes, buildings, things that even the most basic artist would try to straighten with a ruler, are completely wobbly or not connecting. One good example here in the video is the handle of the sword not matching the blade's center, which is a mistake a beginner artist might make, but not the Moebius-level artist the style suggestes, lol.Similarly, circles are never actually circular.
Numbers and Letters are never readable. Unless the image has been retouched, AI models never spit out actually readable text, just noise. So that's another easy tell.
And last one: Is it an elf girl with her boobs out, but rendered in a painterly western style? Probably AI! xD Cheers!
Thank you! Something that bothers me a lot is how peope talk about AI doing what a lot of hobbyist artists would simply consider a shortcut. Asymmetry? Nonsense patterns? Hands? Odd small anatomy details? Ah yes, all things artists famously don't struggle with. There's a reason it's so prevalent with AI, after all.
Some people may have very high skill level, but still do art for fun, so they might not bother with tiny details that aren't the focus of the piece, and just want your eyes to glide over that part. Some artists might focus on the things they personally find fun, especially at the beginning, so you might have an incongruent coloring and anatony, for example, because the artist is more interested in rendering than learning proportions, and yes, it might look goofy sometimes!
The problem is that now you have people who might not know theory (like the idea of just "suggesting detail") hyperfocusing on EVERY part and "calling it out" as a gotcha. They might be right sometimes, but it still sits wrong with me.
Yeah I think a little scripting may have helped here, I kinda just rambled off the cuff and could really have used some more specific language. I'm kinda surprised people are finding the video lmfao, I've worked way harder on things that have gotten waaay less attention. All of your criticism is sound and I appreciate the time you put into this. I will however defend my critique of the skinny boot. It was way too skinny!
@@veno_net It is skinny! It's more that ai, beginner artist or personal style could all make it happen. Anyway, props for putting yourself out there for the youtube algorithm to randomly pick up. Keep up the good fight ✌
@@veno_net I would still say it was a good catch with details leading to nowhere, for example on that sci-fi armored guy. It is true that unskilled or a hobby artist wouldn't give much attention to the details they aren't interested in, however, even if just being unpolished or obscured, those details would still make sense. Most of the time, when you go through an effort to indicate some details, you do that with a reason (think of unskilled artist adding Celtic knots to armor for example, even if they are bad they would make sense in where they appear, or woven symbols in fabric, they may be just indicated, but you as a viewer would still be able to read why they are there). AI doesn't do that because it lacks context behind such kind of details, it just knows they are there sometimes. Also, textures get really weird if you look them in detail, I'm not sure how to explain, but they don't really look like a digital textured brushed ones do, more like artifacts after upscaling.
Well said.
My main takeaway is that AI art makes mistakes that beginner artists would make (EDIT: should say "AI art makes mistakes even beginner artists wouldn't") but the rendering is that of an experienced artist
for now at least. it's getting better and better and in a few years there will be no point paying a real artist at all.
what if I’m really good at rendering, but my anatomy/other aspects of my art are not that great? I tend to have this problem a lot myself, and I’m worried my art looks like AI art.
This is how I used to tell when people were tracing but the other way around - poor rendering and correct proportions.
@@ninibupu art will get stale if no artists can provide new data
@@crowfoot8059then you have to improve your anatomy.
real human art rewards the viewer for appreciating the details, while art generated by AI punishes the viewer for paying attention
I like this point of view
Well said
someone got it right^
How does it punish you for paying attention?
Being forced to comb over these pictures and ask "why is that there" in such a granular way is actually really fun.
Might actually help you appreciate art more. I think we usually don’t spend enough time looking over peoples art and why they put things where they are or why they used this or that color or type of style.
Welcome to the wonderful world of art. This is how people become artists. They start to notice all the little decisions, then they fall in love with making those choices themselves.
@@fallenmango8420 I have developed a reflexive version of this, but with Lego pieces.
I will occasionally just start seeing Lego pieces in stuff as if I'd be trying to build a model of it.m
I have not played with Lego much (not enough at least) for the better part of a decade, but it still happens every now and then.
It's like a hidden object game but the hidden objects are errors.
@@fallenmango8420 this is exactly why i love art. I can think "hm. I wanna see x thing in a drawing. Im gonna do it!" Being able to make desicions, leave little details that most people will miss, do silly things that make no sense out of context like shitpost art, and then follow up with a very thoughout piece... I love that so much.
AI art can NEVER understand this. They just do a whole lot of nothing. No emotion, no thoughts, nothing
I find it hilarious that out of all the material that an AI-made elf girl could gather and collect in an assumingly magical forest, she chose dead leaves. Dead fucking leaves.
She was trying to make real art with them because she's flipping exhausted with AI art. If only she knew...
the melt is a dead giveaway. even if you get lazy the stuff you draw prolly isn't gonna be melted, it'll be messy and ehhh but not... melted. because blurring and blending like that is usually more effort.
also- if you see a detail, ask yourself why someone would logically do that. the giveaways are usually melted and glitchy eyes that aren't symmetrical but not in a way a human would do but in a "this was GENERATED by a computer" way. jewelry and folds on clothes tell a lot. humans usually follow some logic. texture is a giveaway, why would a human do that? would they, would they do that at that level of skill? also lighting- if it follows no sense at all. humans and even animals understand composition and balance. ai doesn't. consistency too.
and when things overlap! usually when you draw for example a hair over an eye you draw the eye fully and match it with the other one and then add the hair and the face below isn't distorted. in ai it gets melted sometimes
and where things go. for example that armor has a bunch of lines and lumps that go nowhere.
Nice video! I like the term "ai spew" for those weird pockets of detailed mishmash. A few other telltale signs I've noticed: buttons and buckles that look like they've "melted", hair strands merging into straps and jewelry, inconsistent pupils, and the art is often really SHINY with blown out highlights.
I call it "Lint".
Meanwhile, Jotaro's hat is fused with his hair.
@@kenicity that one is exception
@@kenicity but everyone knows Jotaro's design is from the 80s/90s long before ai. but thats where the other tale tell signs of ai art comes in if theres AI art of Jotaro. I notice many ai art of copyrighted characters tend to be inaccurate designs despite the level of perfection seen in the ai generated image
The shininess is probably the easiest tell in 90% of cases, seemingly especially when it comes to NSFW art.
Another thing I've noticed with styles with notable outlines is that the outline will have much of the same issues as what you mentioned with hair, where it merges into things it isnt at all supposed to be attached to.
I'm a hobby game dev so I try to keep an eye on new projects being uploaded on itch io, and nowadays the thumbnails on there tend to be infested with good examples of clearly generated images.
And the "games" behind them often just boil down to whats essentially just a gallery of generated NSFW images of popular characters, either in the form of a Renpy visual novel or just a set of sliding puzzles.
Edit: oh also, eyes, these algorithms seem to really dislike eyes, especially if a character is wearing glasses as well.
My art:
- jumbles of nonsense in the background
- poor understanding of cloth
- inconsistent styles in the same painting
- nonsensical hands
…guys I think I’m a robot
....no you arent doing a mismatch
Amen! 😂
at first glance, does your art look like it was done by someone very skilled, and that it would take a very long time to make? The main point of the video isn't to shit on those styles or choices when they are made by artists, it's to point out a larger problem with AI art (lack of intention with what might appear to be high quality art).
@@notreal9214 I understand that, and videos like this are important^^ I was just doing some self deprecating humor
No literally same, like I'm not gonna look up a reference to draw so I'm just gonna suggest some detail in the corner
art prof zoom lecture vibes
AND sound quality
Because AI is not capable of replicating human observation or creativity, it doesn't understand how objects would work in reality. It doesn't understand art theory.
exactly
They also suck at setting. The background can be a quick way to tell if something’s blatantly AI art. AI also sucks at eyes.
that's the whole point. AI doesn't understand anything, it's a fancy calculator
Totally.
So is that enough to stop the economic pressure behind AI image generation?
@@jingbot1071what pressure? companies that want properly done work will hire an artist, companies that want a cheap copycat will hire an algorithm. The only thing being truly f*cked here are indipendent artists getting less commissions
you’re teaching me how to replicate the weird lumpy organic look. thank you. i want that in my own art sometimes. ai art looks like it grew out of a petri dish there’s something so unnerving and bacterial about it. if i can learn to harness that texture in my own artwork id be unstoppable
THAT SOUNDS SO COOL DUDE
Damn you're on point about it looking like a bacterial culture
It actually reminds me of the manga created by Tsutomu Nihei.
Early Midjourney was particularly good at this look
I’d love to see that replicated by a human artist! because a human artist intentionally making their art look that way, is wayyyy better than AI unintentionally getting that kind of look from the sheer amount of pixel distortion from merging hundreds of stolen images together.
I think a big part of the problem is seeing art as a purely utilitarian thing where the results are the only thing that matters. A lot of Ai enthusiasts only seem to care about the immediate gratification of seeing "Sexy Emma Watson dressed as steampunk Ironman ultra realistic 4K" before they move to generate their next super original idea. For me the saddest part is the people that devoid themselves of the satisfaction of learning and appreciating the process and by consequence loose respect for the craft. It´s very difficult for them to grasp that we don't revere artists for making "pretty images" but for finding and developing a voice to make their soul speak in a visual language.
art has always been to me almost more about appreciating the vision, the creativity, the skills etc. of the artist, so seeing people just generating images and finding them so amazing just tells me they don't have any actual appreciation for art. really I personally don't even consider it "art" if it wasn't made by a human. it's just... an image. that's all. and as a result i can't find myself appreciating the images in the slightest outside of "yeah that's pretty, i guess"
@@SanteriP totally
I love your entire comment, and the last couple of sentences in particular. I said this somewhere else but I hope mixed media becomes more prevalent. AI cannot harvest cloth, dried herbs and materials from a trip it took and make something meaningful out of it. Yet that's the art that moved me most
I totally agree with you, well said. Art is about connecting to your soul/the source/universe and creating what you're inspired to create. The process is often more important than the finished piece for the artist. People who appreciate art and craft love watching it being created. Ai doesn't create art, it needs to be separate from true art. Maybe we can call it human made art or soul art so that people understand that it has a soul connection and wasn't created by a robot.
This is something I think it's really hard for non-artists to understand. I can't draw for crap, and I really don't have much interest in it, so AI art becomes a very easy way to jump right into the final product part. As a singer though, I can imagine what artists must think about this stuff because of how I feel about AI music. It's obvious to a creator that the creation process is where the joy is, not so to people who are more strictly consumers.
While its true that "ai spew" is a pretty reliable giveaway, i think its worth mentioning that some artists that prefer to work sketchy tend to end up with similiar "spews" at glance.
I myself tend to create nonsensical designs out of the top of my head in places where i think no attention is needed and space should be filled, just to give a general visual sense of what i have in mind to the viewer. But if you notice common patterns in ai spews then yeah, it becomes obvious, otherwise it might as well be artists sketch work
The difference between you and ai is that the ai is not "working sketchy." Most if not all of the pieces showcased would have taken a human weeks or months, and simply not noticing or caring that a hand is holding a piece of nothing or that an area is filled with pipes that don't go anywhere ceases to be a reasonable explanation.
I found the first piece he showed the most believable, because at a glance it looked like someone throwing flats and subtle shading onto an otherwise very sketchy piece. The fact that most of the nothing is supposed to be ruins also helps. I feel like the sword hilt and the boot are the only real givaways (the best givaway is the bit of sword hilt on his arm).
so your saying your drawing a fucked up hands
@@hello-xm5il artists nowadays have a lot of tools to create details. We don't live in 18th century
i wouldn't say so, it's obvious when it's a sketch or mistake or laziness or human error or action- what humans do is usually intentional. varying levels of intentional but definitely still lucid unless you're on complete autopilot or not sober. what ai does is clearly unintentional and random, melted, fuzzy. complete nonsense that is easy to notice if you train your eye.
hell i've drawn completely drunk and it looks better than whatever ai is doing
I think one of the most effective ways to recognize AI "art" is simply to expose yourself to it. Look at it enough and you wil lstart to see patterns, shapes and colors that make it almost immideatly obvious you are dealing with a generated image. Be it realistic, stylized ect theres always indicators in all of them that will stick out to you the more you see them.
another way to tell is if it has shad's wife's head crudely photoshopped onto it
@@numberonedad 😂😂😂
Even better. Try it yourself.
My brother who's obsessed with rats was making a ton of ai rat images on Midjourny. In ones where he tried making crowds of rats you'll see them get copied and pasted and melt together into rat soup.
He let me try my own promts and as a curious artist. I wanted to see mythical creatures that rarely ever get drawn. I'll describe them the best I can and the Ai wouldn't even get close to it and would even make random stuff that has nothing to do with the prompt at all.
After that I felt pretty secured that the stuff I'm most passionate about drawing can't be replicated by ai.
No not really. You're only thinking of stereotypical AI art used as examples of bad AI art.
@@Nurariartthe problem is the people giving prompts and instructions have no experience or knowledge like you and your brother
AI art is kind of like how you imagine a very complicated art piece in your head
when you just imagine it at a first glance it looks so cool and amazing
but when you try to flesh out the details of how would it look like exactly, everything just breaks
That is a profound and yet on the money way to describe it 👏🏽
I think the main takeaway here and why so many people fail to recognize AI "art" is the fact that most people don't know how to truly appreciate art. They will see a piece of AI art online, or any art for that matter, and look at it for maybe 5 seconds, think "wow that's so realistic! I could never do that!" and move on. Few people actually stop to take in the details and appreciate the time, skill, and effort it takes to create a piece.
I wish more people would treat art in online spaces the way they do in an art gallery - not rushing past every painting on the wall, but really stopping and staring and noticing.
I love pepperoni pizza
Is that necessarily bad tho, for most ppl art is a decoration and doesn't have a meaning apart of aesthetics. Like if you look at a curtain can you see a deeper idea in the floral pattern, no. Not everyone cares about art and it's not a wrong thing to do.
@@AGM_V2I think that is kind of wrong, actually. consumption for consumption's sake, whether it be AI art or floral patterns on curtains or endlessly scrolling through tiktok is a bad habit to be in, albeit to different degrees. i do think it's been normalized, but in the same way that alcohol and cigarettes and fast fashion have. we've gotten very used to this shallow, aesthetic relationship with art and I think that's a lot of the reason our media, our clothes, and our furniture is cheap garbage (doesn't help that the material stuff also produces lots of waste!). it doesn't have any real purpose besides existing for the sake of it. not to be seen, felt, or understood, just noise.
@@katto4073 yeah, but consumins is good and you can't change it, ai is just next step to it
this is totally right on the mark. I find myself missing the details that would clue to a work being AI art, and it's always because I just glanced rather than actually taking in the work
They're not scared. They're lazy.
They don't want to put in the effort, and they're satisfied with the fantasy that AI fulfills for them.
My opinion as an artist: what's going to end up happening is that the bar will be raised. That "caliber" of AI art is going to become the starting line and real artists are going to aim to surpass that bar. Artists will focus more on illustrations with a story--with context, theme, and narrative--and work even more to build a personal style with a solid reputation.
They will also focus on designs with authentic organic patterns -- the kind that is very hard for AI to interpret.
Personally, I'm not a doomer about it. The real advantage AI has is speed and output. But until they create an AI with the same capability as the human brain, a real artist is always going to be more skilled.
Your standards are valid but close minded, because others can express themselves differently you might want to take something away from them?
Would you agree that there are monster artist out there that set a standard so high it's almost untouchable? Are they "Doomers"? Did they steal all Art from you?
No, Art is the size of of the universe, we have many many worlds to explore, it shouldn't be about conquering, it should be about exploring.
[Why do I sense you're hinting at financial gain instead of the love of art]
@@TheManOfReason.This ai are trained by artist'sbdrawing and creation
@@Hahshdhbcbcyoutube And how do you think any actual artist starts learning ? They would also look at others art and take inspiration...hell, sometimes imitate things they may like as well in their own art. Everything anyone creates is inspired by what they have seen in like which guess what ? Also includes art of other artists.
@@amadeusnagamine9056 I'm so tired of this "but artist use inspiration just like ai do! Bla bla bla " Well here's the problem, ai can make a perfect replica and humans just simply can't. No begginer takes, for example samdoesarts work and make a perfect copy of it. Please watch more video about this topic instead of bringing the same thing to the table
@@Hahshdhbcbcyoutube
*"ai can make a perfect replica and humans just simply can't"*
As of your knowledge, I'll grant you that it maybe the case as of yet, but not in the future. Humans are amazing so don't hastily steal away what we might know yet.
*"No begginer takes, for example samdoesarts work and make a perfect copy of it"*
Agreed, give them time and would you agree they could? So time is only an issue here?
So my understanding what you may be trying to express is:
The time it takes for A.I vs Human to create/learn how to do a piece of Art.
My advise if it's your concern, is everyone and everything that can learn, learn at different speeds, even now A.I. is "learning" and improving and adapting. My point above is, what is your goal my friend? Is it to conquer or is it to explore.
(One want's it now, and the other is timeless.)
So? To conquer or is it to explore? (or something else?)
"A detailed piece of nothing" is probably the most accurate description of AI art that I've encountered
I’m not a professional artist, but I draw for fun. When I see AI generated art it gives me an uncanny valley feeling. At first glance it looks good but SOMETHING always feels….off.
As someone that does this for fun, I like the idea of AI helping me get the repetitive work done, or helping me do the time consuming stuff.
For the first one, it wasn’t the “spew” of the AI that tipped me off immediately, since that particular piece was clearly generated using Mobius as a prompt- and his style is known for that sort of blending of organic shapes with technology. What jumped out to me before you even zoomed in was the weird perspective and angling of the sword. The handle is crooked, like you pointed out but also looking at where the tip rests on the ground, it’s just not quite right for how it should sit in a physical space.
"rests"? that shit is BURIED
"AI SPEW" is such a great way to describe that whole detail creating nothing tendency that AI images have. If you can look at some bit of detail and go "yeah, but what is it supposed to be?" that's usually a good indication of AI art. There's lots of other indications that I come across that are obviously AI, but a lot of those feel barely worth mentioning because those will get better in time, but the things I think will NOT get in better in time (or at least not as quickly) is something that mainly accomplished real artists will pick up on. It's when the image is rendered to a high degree, but also wrong in some fundamental ways, and any professioonal level artist will instantly see that any human that could paint or render to that degree would not still be making those kinds of fundamental errors in light sources matching, or perspective, or anatomy, or a thousand other tells. "AI artists" who claim that "people can't tell the difference" just have acase of not knowing what they don't know, because sure THEY can't tell the difference between human art and AI images, but certainly accomplished artists who have seen enough AI art can see it pretty clearly.
"any human that could paint or render to that degree would not still be making those kinds of fundamental errors"
Yeah, I think this is the biggest tell. Simply the distribution of errors is nothing like the distribution of errors in a human drawing. Sure, all those mistakes that AI makes could've been made by a human, but the overall combination of "awesome" and "awesomly stupid" in a single illustration is different when done by AI vs when done by huan.
the sheer IRONY of putting a watermark on AI xD
AI can now replicate text convincingly well under the proper conditions. It now understands how hands and feet are supposed to look. It can tell the difference between hair and the environment. But it can never replace human creativity. Poses are something it seems to still struggle with, especially very dynamic or action oriented poses.
Ai struggling with the same things as me fr
So what make you think it won't learn?
@@ghostoflazlomost people are bad at using prompts and don't use photoshop to touch up flaws in their generations. Drawing something gives you full control over your final result instead of being forced to correct things and generate stuff constantly. AI will always fuck up something in the process
Let me know when you can provide a rigorous definition of creativity and I'll consider caring.
@@mrosskne come up with a rigorous definition of... my nuts
3:29 One little thing I want to add to this notion about how a real artist wouldn't make that kind of error- It should be specified that an artist of that particular skill level wouldn't make that kind of error! Generally, you need to think about how the presented errors compare to what skill level the piece would represent had it been made by a real human. If the art and design appear to be of professional quality and skill but has errors like this, then yeah, something's definitely fishy there. But if the art looks to be made by someone who not nearly as skilled, then it would more likely just be a matter of an error made by someone who's simply still learning.
Of course, it's still very sad that we even have to scrutinize any good-looking art these days so we don't accidentally support lazy AI users who think messing with prompts takes the same kind of effort and skill as making genuine art.
And I hate how a lot of people call themselves AI artists. There is no such thing as AI artist, what is artist about it, you write a prompt and you're the next Van Gogh?!
@@Renikee Yeah there is literally zero difference between an "AI artist" and someone just googling "cool space drawing" and finding their favourite image. They have put in the exact same amount of creative input, it's just one extra level separated from the original creator.
The problem is the exact same type of person who thinks AI images are art that they created, also have never experienced the drive to create art in their lives so don't actually even know what it feels like. They describe themselves as artists, but they actually don't actually understand what that means.
also of note: ai images are usually have widths and heights of 1024 or 2048 pixels
I hope there’s more regulation out around what kinds of images can be fed into ai image generators in the future. The biggest fear for me isn’t so much that ai art can look like human art, but rather that most of these image generating softwares like Dall E and Midjourny use an unethical data scrubbing system that takes data from artists online without permission or compensation. It’s unbelievably exploitative.
I hope the current lawsuit against Microsoft and OpenAI will have a domino effect on all of these other AI models. Maybe the datasets used to train AI should be limited to what is available in the public domain/creative commons or at least purchasable from stock media sources.
@@enviritas9498 it would just make the AI sound like its from 1850s. It would really look good. Would you want to stop progress just for that? What about the science?
There was a post of conversation with the midjourney creators and they had a whole list of artist they stole from while encouraging additional names to the list. Meanwhile theyre claiming to media they dont know who they stole from.
@@TheManinBlack9054 What kind of progress does AI image generation help with? What science? Art is a creative outlet, it's inherently human. Take away the human and it becomes nothing more than a mesh of patterns and color.
@@TheManinBlack9054 plagiarism from artists isnt of any value at all. can't say i feel any different when its techbros.
One of the things I notice the most is that because image generation models make everything in a shotgun blast of everything all at once, it has a similar effect to a person creating it with immediate amnesia of why it started drawing any particular line. They end up continuing and sharing lines between different component pieces.
So you end up with these weird lines that start definitely as part of one element, that just all of a sudden become definitely part of another. The model has no idea that these are two elements, it just knows "lines continue" and "line goes here" - how that interacts with the rest of the composition doesn't weigh in to its decision making. From the model's perspective, it has correctly fulfilled the request of having each pixel or group of pixels make contextual sense with the ones next to it.
You can see it happening with the 1st image's right boot. There's a line that starts definitely as a fold in the leather, that suddenly becomes definitely a detail of the ground between his feet.
The 2nd elf woman's basket has lines that start definitely as the rolled edge, that suddenly become definitely strands of twine.
That famous AI art of a buff Trump, where his legs and pants just blend into one another.
I saw a video where an expert was explaining why a faked photo of a supposed armoury generated things such as rifles with barrels coming out of both ends of s stock, or multiple magazines sticking out of it: the software knew a barrel and magazine was supposed to come out of a rifle's stock, but "forgot" that it had them already coming out one side and so cheerfully kept producing them.
I love this. As an illustrator myself, I always found it relatively easy to tell when art is AI generated, almost like second nature, but I know that for a lot of people that’s not the case. I think this video did an amazing job explaining the logic of it all, and put into words very clearly an instinct a lot of artists might already have when analyzing if an art is AI. I hope this video blows up so more people can also learn how to differentiate AI from real, genuine human-made art.
I was not prepared to grapple with the fact that 2018 was 6 years ago 😭
Oh damn… wow. I mean wow what wow.
Yeah initially people were amazed by ai art. Now it's like it's hard to imagine ai art being used anywhere legitimately because everyone knows how to notice these "suggestions of something". And once you notice just one of these in a setting, it's a turn-off.
Quite a few companies have tried...
For example, I still have no idea how Wacom of all companies, an ART based one at that! Posted Ai images and used the excuse "We didn't notice it was ai" When the detached dragon tails and fuked up teeth were sooo obvious. Lol
Sadly i'm not that sure about this.. I see lots of book covers in retail that are obviously AI Generated yet it doesn't seem to bother most of people (and certainly not publishers c: ). Same for birthday cards or whatnot. Lately Palworld a game that is really popular on steam (sigh) is a mishmash of asset flipping and IA Generated spew. Lots of people don't have the visual baggage to differenciate the AI spew from real art.
I wish people would notice and be turned off but i'm not really convinced it will work that way sadly :/
There's AI art winning art contests mixed in with traditional art. Also this is AI's early years while humans have had millions of years to brush up their game
Most people definitely cannot tell the difference, even past a cursory glance.
@@karrawr9538AI’s artistic limit is about wherever the current limit of humanity’s skill is wherever the time period it’s in as it’s only capability is to mix, mash, and combine already existing styles into an image with zero personality. The only real thing it can do as of currently is to refine rough edges and integrate prompting more accurately, other than that, it ain’t going far.
"a suggestion of something" is exactly the way to describe it.
I guess we all have to slap buttons buckles and belts on characters like it's the 80s again XD
Rob Liefeld vs the machines
Yes. That.@@danielwesley5051
Hell yeah! Let's reboot The Pouch comics 😂
That as well. Heh@@oliverlee4033
Holy crap! NOMURA KNEW!
I think AI art is just like fast food , it only appeals to people who wants the dopamine rush to look at something for a small amount of time which is one of the reasons why AI art is so popular in social media sites like instagram . Only few people can enjoy art and spend time collecting and provide valubale suggestions for them .
Nah its more like matrix nutrition paste imitates glucose and protein but has synthetic snot taste but they trick themselves into thinking its mcdonald food then its not even that
i see your point, but unlike art, food is a necessity and some paychecks don't allow for good quality food so that doesn't really equate (lighthearted). can't say this is a much better analogy but i'd compare it to scrolling on whatever youtube's version of an fyp is vs watching long-form videos. you'd rather get handed whatever the algorithm thinks you want as quickly as possible regardless of quality than search for a creator you like or a topic you actually want to engage in.
AI will have that "too good to be real" look from a far but up close it has this unclear, foggy, and sometimes weird look to it that's super common among AI images (especially when it comes to humans). I am really curious though how exactly AI produces art like this.
For some reason the low audio quality feels soothing in this video, I like it.
it’s crazy how whenever i try to search google for pictures of any artistic sort, 90% of it is AI generated. it’s really getting on my nerves, and as much as i was thrilled about the usefulness and capability of AI at first, i now started to hate it more and more.
the good side is that will make AI worse, because it steal photos from internet, and most of these photos being AI... well, i think you can understand what will happen.
(sorry for the bad english)
Im an artist as well and I have tons of issues with AI art, some of which youve mentioned in the video, but one of the biggest and most saddening things i hate about AI art is it has forever changed how we can view and enjoy art.
We can never just see a beautiful artwork and immediately be in awe of the effort and skills of the creator.
I remember i used to love just going on Pinterest because I loved seeing tons of amazing, incredible art and adding them to my boards which would continually grow as i continue to be suggested more and more beautiful artworks. Now Pinterest is just filled with AI garbage and is just a site that reminds me of the exploitative technology. It's like visiting your favorite beach and suddenly realising the water is full of plastic and trash...
Totally agree. I do use those AI arts from pinterest as reference and use my own style and stuff i use and make my own. Im stealing from the stealers >:)
Thank you so much. I'm not a digital painter and not much of an artist anyway, but its nice to know how you recognise it. To be honest, in many cases I would've been fooled... I would've made excuses for the artist, saying "he didnt know better" or "he must've thought something, I'm just not seeing it" ... I still find it hard but this certainly is a start.
Bros mic is ai generated
On the Sci-Fi/Fantasy art- I seen a guy popp off, a guy who from his writing seems to be quie a good writter, but they been using AI for illustrations and you could notice the "badass art about nothing" vibe you were talking about. Plus most of the "art" output looked like if it was mildly trained on Fallout concept arts.
I was mostly disapointed in the guy, could have scribbled things themselves or get someone to do it for them- and the "I ain't have the time" is a fkin insulting cop-out
Why is it a cop-out? I don't see a good way to achieve his illustrations. His own scribbles would prob. have looked horrrible and put people off. Finding someone to do it for you for free is quite rare, you would likely have to pay for the commission. Apart from the fact that the model was probably trained on images their artist have not given consent to (which is immoral), his approach seems pragmatic from another writer's perspective.
you all did not care when they used scraped writing to train translators and AI text generation. Why should writers care about your art in return?
One of (edit: The only) beneficial things of AI people posting a zillion images a day is that it's kind of self destructive to generative AI. Since theres sadly no real reliable way for an AI to look at a billions of images and categorize them as AI or not, when they go to scrape and steal images for training data they'll inevitably end up with tons of AI crap in there. (And much of it not tagged as AI by the poster because many AI people dont want to open and honest a computer made it)
Meaning errors or bias that propagate in the output might be magnified because they're now present in the input. Not sure how these AI companies will deal with that because it's a problem for them that'll only get worse with each passing day.
So don't worry about AI replacement of art; it isn't happening.
we already have AI that can identify generated images with >95% accuracy
AIornot, isitAI, hive AI detector, maybe's AI art detector, etc. This isn't new.
That's good to hear
i draw in a quite realistic style with some elements of abstraction (a lot of randomized pencils without necessarily drawing a sketch or specifying what it is) and i get often accused of being AI, it's really annoying. The AI copied me, not the other way around.
got any social media you'd be comfortable with sharing? I'd love to see some of your work
same! id like to see your work
@@ceilingfanvi can't post it for sone reason my comments get deleted
@@augustopenaspalmeira471 Just tell us your username
@@augustopenaspalmeira471 youtube doesn't let you post links, you have to post a username or something
Likely the greatest guide on how to differentiate AI art from human art ever created. And the best RUclips video ever created too
The immense output is a good way to spot an AI prompter, as well as checking the art style. A person's art style will evolve naturally over time, but if several pieces posted within days of each other each look radically different, that's a red flag. (These are the same tips commonly used to catch art thieves, which, let's be real, AI "artists" essentially are. Some things never change lol)
I feel like the flood of IA generation we're getting is going to loose it's worth quite fast in the long run. Those people generating art assume that the human eye cannot be trained to recognize what's human from what is not. Fun fact: it is. Just like with any new concept we discover, it takes time for us to "adapt and get used to it". But after you understand the general parterns and repetitions of what an IA technically capable of, (by also deepening your culture and knowledge allowing you to recognize which human artist's work got fed into the machine) then it becomes insanely obvious and jumps to your eyes the second you check those "art accounts" :)
It's also gonna lose it's worth in the long run due to accidentalyl feeding AI generated images back into the training data, which reduces the output quality
It's AI. Not IA.
sorry to be french @@mrosskne
@@EkattPalladium The rest of your post is in English, and you're commenting on an English video that uses the English acronym. Don't pretend to be ignorant.
@@mrosskne bro I just exchanged two letters without meaning to because I'm so used to say "intelligence artificielle". Chill
Another thing that generally stands out to me in AI art is just the choice of color, usually. Generally it looks very uniform and uses the same sort of dull, desaturated colors. It may be purely anecdotal on my part, but I always take a look at the color. Light sources, too.
To preface this: I am not an artist, the most I've done art wise is just the funny channel avatar and my profile banner. Not particularly great work (especially since the cropping on channel banners made it hard to include the full piece that should be there). But it's still possible to notice these details if you have some clue as to what you're looking for.
No I am an artist and I pay attention to this stuff and I can’t tell you because I’m deep in this I can tell you artist that I study by name I have a whole Pinterest dedicated to art and things I want to buy and other things I’m interested in and ai can create very vibrant pieces of art the best way to tell something is ai is the eyes of the characters are very fixated and the faces
I really enjoy what you call the “AI spew”, when you have a hint of something that turns out to be an illusion, this is what I strive for in my traditional painting.
i love how tip #2 is basically just "spend more than 5 seconds looking at the art and also maybe even go so far as to think about it as well."
Your microphone gives me nostalgia for the moon landing.
Commenting to boost the algorithm, this video was very good !! Telling apart AI from real art is super important, definitely sharing this with people to spread the info. Sending all my hugs and good vibes to you !!
Also, ive seen some people defending AI generated images as "art" by saying its "making art accessible" as if theres something stopping them from picking up a pencil and a piece of paper ತ_ತ they act as if art and creating is like a coveted craft only certain "talented" people can do and its really just sad, because it shows how little will to practice and actually care about what they do they have. Anyway, just a little rant from me lololol :)
I fear that a lot of up and coming artist, myself included, face the hardships of dealing with AI in the industry/ art space. Already don't like that fact that discerning what is AI and what is actual art is a problem especially for someone who is not well known. Even worse when you see a professional concept artist start to use midjourney, then it comes into question who is mixing it in their work or not. It's such a mess. Great video btw and hope you continue your art journey!
gooners really have no imagination, that's often the giveaway
I mean a lot of artists have no imagination, go only literally and art site sort by new with the AI filter on and 95 percent of what you see is bad art and uncreative compositions. People like to compare AI art to the top like 5 percent of artists and not the sea of absolute schlock that people actually tend to put out.
whatever helps you cope :)
@@f556784q3 L take
@@f556784q3I think this is cope. Even if it may not be some Picaso or something, at least those ppl actually take the time do the thing themselves. Every artist starts somewhere and shows some drive and passion.
"Well if you think thats bad this is worse! They lack imagination so that means I do the same thing but with less effort by having something else do it for me and only works because of said artist!" . It's playing the mediocrity game.
True. Ai "art" can generate some really amazing weird stuff if you let it. But most of the people using it generate the same generic images of hot fantasy girls
people think ai is a shortcut to cool art. they get a rush when they can instantly manufacture a flashy image. but that's all they get. a rush. no greater understanding of their subject or how they relate to it.
The way i could tell whether its ai or not is when there are ugly random glitch colors and and the ears with the earings and the weird clumbs in hair (which is usually FILLED with hair strands) i dont really have a problem with ai when its not used with ill intentions like a paying customer for a comission. Some are great for giving ideas
I feel really sorry for artists. I once dreamed of being an artist, of learning how to digital paint beautiful fantasy art like frazetta, brom and guys like that, those D&D artworks stuff, also scify... all the beautiful things they were doing out of their uncanny talent they built over decades of studying.
And now their work have just become replicable by ai in just a feel prompts and a couple of minutes.
I feel really sorry for that. Had I chosen this path for my life, had I followed my dream, I would be devastated now.
You know in a way, i think AI might make the art community stronger in the long run. Right now we are in the thick of it, the chaos days of early AI havoc, but i think a stronger vision and respect regarding what art is may come out of it. This may take time though, years.
In a way, there is huge engagement in art right now, all these people doing AI and people talking about it. They will not keep doing AI. But those who stay after the fad dies out, realizing they haven't built up anything in their years using AI, may start doing actual art.
ultimately it depends on the people. the creators of midjourney have been caught with a discord screenshot proving they're plotting to launder money and dehumanizing artists (if you need the link i can't post it because youtube doesnt allow links i think) but im sure you can find it somewhere. im sure ai art will help a lot but right now yeah it's a cesspit
@@lemonadesaccounttm8721to think that one company defines a technology is a bit stupid, don't you think?
@@lemonadesaccounttm8721 wow that's messed up. But nah i don't need to see it. It's very much about the people yes. There will always be bad players. But the good ones will prevail!
Copium
@@itsgonnabeanaurfrommeI wouldn’t call it entirely stupid. Google became such a phenomenon that it became a verb. Amazon also defined online shipping sites. They are some of the biggest, but partially because they were some of the first. With the world we live in, a company can absolutely be representative of a technology, especially in the short term, where they are one of the only few ones widely supplying it.
Besides, the replier had a point. Midjourney is one of many AI ‘art’ programs whose users consider art nothing but a product, and right now, it is a cesspit. And I agree with both the replier and the original commenter, that it might have future uses and be a net positive. But for now, it’s causing harm, and Midjourney is just one example of that.
From my observations, AI art usually struggles with consistent lighting. Primary light sources are affected by bias from the input data, the light source can be behind the character\object, but the front will still be lit because artists prefer to position their light in certain ways. Shadows are guessed\approximate and not consciously projected. And bounced light usually incorrectly represents how the light would interact with the surfaces in the scene. There are some ways to improve it, for example using a simple 3d scene from Blender as a rough input but the result would still be guessed based on the collected patterns and biases because that's how neural networks process information.
“It loves art about nothing” is a perfect description of AI-generated stuff 👌 That’s exactly the feeling I get when I see it, and it immediately signals to me that no real human was behind it (except for the artists that the AI was trained on).
Thanks for posting this. It needed to be said, and it will need to be said many times over. I'm a non-pro digital artist who spent years building up my own style, that sadly is very wispy/dreamlike, and therefore quite easy to be drowned out by AI. I got pretty despondent about it at first, but I've since turned it around, gone back to learning how to tighten my compositions, but also how to use oil and other physical media to augment my work. When I start posting again I will include images showing my rough work and progress to prove that it's my work.
I totally agree with you that AI cheapens art, and that people who use it don't seem to realize that art is as much the process of creation as it is the result. But I think it will cause people who truly love art, artist and fans alike, to appreciate that work more, and to value and support people who are doing it for real.
Another thing I’ve noticed, usually with AI pictures that have multiple characters, is that the lighting and color between each character and the background would just be slightly but noticeably off, like a bad green screen where the lighting on one thing is very different than another.
I draw my more detailed scenes like the first ai lol. I work off a lot of random shapes to imply things and just hope that people don't look too hard into it. It's one of the reasons I generally don't draw whole scenes.
I also am just awful at planning so that's probably another reason I throw random stuff in lol
One of the saddest things about AI for me is I can no longer just enjoy art, when scrolling through pinterest looking at cool art I'll always have to be suspicious of everything
Before all this I could just scroll and be joyful at others creations, now I use pinterest and just feel empty
Mental issue.
He's basically roasting AI-Art and I love it hahaha
My biggest giveaway is the fact AI art looks too 3D and polished. Another big giveaway for me is the mistakes AI makes sometimes. Eg- 6 fingers, 3 nostrils, jumbled up text like "frnczon 2"
this helps explain why when i see AI "art" I feel as though I'm gazing into the void
I really love your take on this and you put it very eloquently when you said that AI "art" is made by people afraid of trying, afraid of failing, afraid of being engaged in the art process. It is a sad day we live in as artists and I worry for the future of artistic integrity. I think traditional media will have more value in the future as digital art becomes more and more saturated with ai generation. The frustrating and sad thing is that we are increasingly seeing that the people who care about art are artists. Which is to say that the majority of the populace just care about the picture they look at, not the way it was made. They want to see something pretty (hence all the subjects of ai art being pretty women/men in popular themes like cyberpunk) but don't care how it was made.
AI doesn’t draw from fundamentals first, it renders out what it thinks something is
I'm so tired of having to analize every pixel of any art i see online before sharing it.
Good vid. It's not an easy time to be a learning artist so just keep at it bud. Good on you to fight the good fight.
Seeing this video pop up on my youtube feed was nice, because I see so much AI art on youtube and other places, it makes me mad. Thanks for helping me identify weather it's real or not. I respect the passion and time that goes into making real art.
You've obviously put a lot of thought into this, I like the thing you said about AI praying on people's disinterest. It was insightful.
Good indicators are also "AI scribble" where a bunch of letters got layered over another to create the illusion of text, additional comes "AI grain abstinence" which is basically just the absence of luminecant noise and color noise, when you turn down these two facors to a max in Photoshop your handmade picture will look more like made by AI, the very smooth "non-texture" is rather characteristic for AI imo
this guy definitely plays melee
Lmao
You're so good at pacing and speaking in an engaging way that when I started this video I was like aw man 18 minutes I'm not gonna sit through that let me just add it to my watch later but I started to watch a bit just to get a taste and suddenly the video was over and it felt like no time at all.
"Hello, this is mission control." Ahh microphone
I actuallt tried generating some ai art(not posting of course) just to see if I could notice the details, and the mismatched style is actually such a large giveaway, it was really blocky and suddenly got photorealistic for the skin, this is super helpful
I'd be very keen on seeing you do a blind Spot-the-AI video. Say 7 images, chosen by another artist, and you examine them to pick the fakes and see how well you do.
I once even made an acrylic painting of pokemon Tauros inspired by Van Gogh painting called Lying Cow and I wrote "to prove it's not AI I'm add the picture of my work in progress painting". I think it's important to add work w.i.p. for some art to show people that we not make an AI art.
a lot of ai artist also chose topics that are very common for beginers
elf lady is legit the firs thing i tryed to draw when i was 9 and decided to learn and i have multiple drawing from elementry school with the exact same posing as the guy with the sword it's just a little funny to me how all beginers seem to have the same ideas of what you shoud make
I like to look at the hair, eyes and detailed patterns. Usually strands of hair will mesh together in ways that don't make sense.
Great video. I think it's sums up many ways to find it's ai. Probably the biggest point is indeed checking the socials of who did it
This video just put into words a lot of thoughts and feelings that I've had for AI "art" for a long time. LIke the blending of styles, and the inconsistency of photorealism and surrealism in one picture is something humans would rarely if ever do.
the AI watching this video like "Thanks Prof! I'll try to get better by next time."
That's not really how it works lol
You know how a joke work ?@@sarabeatriz5569
This isn’t something that’s easily fixable
Someone said that “think as an artist”, see any detail and ask “if I’m an artist, why did i put it there, why did I draw this detail” after that it is easy to see ai art
The example at 7:09 is just wild to me. I feel like a crazy person telling people just how CLOSE some of these outputs are to the original art and/or each other, NONE of these GANs are producing 'original' work, its pieces and bits closer or far away from a small pool of originals that it grabs within a processing loop. There's really NOTHING to be 'impressed" about here other than it's 'interesting' how this method to processing iterations of pieces between bits of noise and grain are how you get a derivative result, and the speed in which it's processed.
IE: NO ONE is impressed by taking a photo and putting photoshop's "Oil painting" filter on it, it looks like shit. But essentially, that's what GANs do, just in a more 'seamless' way, by setting up a 'composition' that's practically a photobash, and adding or removing 'eased in' elements from other photos either a little or a lot. Just like the photoshop filter, this is done through using the noise that is present in all images through the color data / pixels
The fact people genuinely believe GAN assisted imagery is created from scratch, pixel by pixel is so ridiculous to me.
Rant aside, awesome video. Doing my part in watching and commenting. Just passing through.
It would be interesting if you could make a video guessing whether various drawings are real or ai generated
I don't think AI will ever surpass humans due to being bound by them. Mimicry is a very accurate way to describe it.
it already did. Majority of humans can't draw something comparable to what AI can do, even if the AI art is flawed
@@cccbbbccc5910 Rendering is the primary metric most non-artists use to judge art, but creativity is rarely considered.
Think of a prolific artist or studio with a signature style who spawns dozens or hundreds of imitators. We can easily spot the influences of these artists and recognize them as imitations.
Most of these imitators are unremarkable and don't progress the industry. The element that makes an artist stand out among a sea of imitators is what AI will never have: creativity.
@@Tubeytime I'd argue AI can be creative. Unless you have some romanticized view of creativity
I saw that Beat from JSRF in your gallery while you scrolled, sign of a true artist 😎
When I draw or watch artists, a lot of them do like beauty, but, there is always method, usually. Like, the folds of the hair and clothes usually have a pattern to them, and a method to them. Like, Shinichi Sakamoto loves to draw bishoenen and biseinen, but there is a lot of method to his characters.
I've always been a huge fan of weird or horror type of art. Unnatural stuff that most people would never think of. I love some of the tnings ai spits out sometimes. Like the weird optical illusions. It looks right but something is off. Makes you feel uneasy. Legs turning into arms. Small details that sometimes are so creative in a way that a human would never think of. Especially if youre making pornographic stuff but then suddenly one of the pictures is just creepy and horrific. Always been into fantasy art and horror art. Disturbing stuff reallytaps into your emotional responses. Ai does this so well. Its mind blowing.
Keep in mind, There are alot of false positives. There is currently no relaible way to determine if art is AI, and beginner artists can make alot of these mistakes. (I know I do) Drawing pretty woman, Having bad folds in clothing, and having a boot that is too small, having an inconsistent style, are all things that I would do lmao, and I have been drawing for ten years.
Also, AI art in skilled AI creators (those that use control net, photo editing, LORA, ect) you wont really be able to tell.
im definitely guilty of the ‘pretty women’ thing too, but ai doesn’t it in a very specific way, usually a painterly anime-esque style with fantasy elements
@@elvenbugs one of my favorite artists, wlop (the creator of the knight shielding princess meme) does it in exactly the AI art style. Kooleen does this too.
You perfectly explained why AI generated images are so uncanny, the fact that it generates organic looking globs of nothing that gets more frustrating to try to figure out the longer you look at it. Because it looks so familiar but at the same time makes no physical sense. Even in the setting of a fictional world the background usually starts breaking the farther back it goes. It’s incredible technology don’t get me wrong and we’ve already been using a form of it for years with photoshop but at least there’s still work and effort and style put in. I really think that AI generated images should at least have a watermark or sign saying that it’s AI, because a lot of people are fooled easily.
Before I start and for context's sake I would like to emphasise the fact that I am a computer scientist who dealt with AI on a foundational level (not just prompts).
I see all this well put critique as valid at the time of your recording. Thank you, I've subbed.
IMHO it will only be a matter of time until things will get sorted out on the AI generative front, making it far harder for humans to tell the difference of what's men made and what's not.
I don't share the optimistic opinion that good artists will leverage these algorithms to boost their quality, throughput and reach. Proof can be found in the content of your video itself. These tools are still way too fresh for the collective consciousness to have them absorbed and know what to do with them (and, arguably more important, what not to). Psychologically speaking, this does explain why there seams to be a a natural incentive (call it hype, if you will) where more and more people will try to cheaply capitalise on these low-effort cash-cows that are art works.
Instead, I do believe that, for a time at least, human-made art volumes will go down (to what extent, I don't know). Many artists (known and unknown) will feel the burn and will probably give up the field altogether.
It's not all doom and gloom, though. Art constantly needs to refresh itself all the time to stay relevant and have an appeal in the human psyche. Having no real art to feed the machine will inevitably lead to a degrade in quality of the AI's generated content. This will get noticed and affect its viewership which will again allow for humans to come and reseed the field. How that artist will look then it's anyone's guess. I do have a hunch (or rather a hope) that when it will re-surface, it will contain some sorts of nostalgic vibes woven into it.
Of course, if we will manage to build machines that physically experience the real world before that tipping point will be reached, I do think this causal trajectory would have to be revisited. If this does happen, though (can't tell the odds, but they're not improbable), and machines will be able to generate content based on experiential input, we would find ourselves in a totally different philosophical realm of existence that would weight heavily in favor of artificial consciousness being achieved.
The series "Altered Carbon" portraits a really cool concept of a future in which humanity avoids AI just because it knows it can't escape it's optimised purpose of existence even though, ironically, humans were the ones that brought them into existence. A likely pandora's box ending scenario if you ask me.
Thank you once again. Take care!
The best thing to come from this is I get to watch more paintover-critique style videos that artists are too afraid to do to other human artists
Sorry, but I just need to say, I feel like saying AI art preys on people's apathy toward learning how the world works is a little harsh. It's pretty obvious to me that the reason AI art is so prevalent is much simpler than that: it is designed for the art sensibilities of social media. That is to say, it is easily consumable and looks passable at a glance, so most people won't investigate beyond that. You could call this a kind of apathy, yes, but only the same apathy as we feel toward most other internet "content" - that is, an apathy which is merely a symptom of the modern age and, in my opinion, does not reflect poorly on a consumer's character or their interest in the world around them. I am certain someone could be familiar with how clothing or mechanics work and still miss some of these giveaways that seem very obvious to you as a digital painter.
I saw you state in another comment that this was off the cuff and could have been worded better if scripted, and that's fine. I'm not asking anything of you with this comment. I just thought it was worth saying.
Yeah, i agree with you!
It very much feels defensive.
But can you deny the world general population is becoming apathetic, even in real life? I would agree that the apathy is here long before AI Art, but wouldn't AI art feed into that apathy and encourage it to grow? Would AI arts inspired anyone to be an artist, in the same way a bland yet beautiful anime art style did? Is AI Art making effort seems worthless? We used to be inspired by one another artist and have a discussion about it, now what kind of discussion will we have instead?
What will happen to the next generation? The one who wouldn't experience the world before AI Art?
This is about iPad kids with worse fine motor muscle, who can't write well, can't focus well, can't communicate well.
@@EarthWingedDragon Congratulations, you now feel like everyone in every other industry.
"What will the next generation do?" Torch a lobbyist or two if they know what's good for them.
The keyword for Ai check is 'Intention' AI art has no intention.
Not sure if it’s in all cases, but AI can’t generate hands or feet properly. Sometimes there’s six fingers, sometimes four 😂.. so that’s one fast way to tell.
It's largely based on the model being used. The people who actually put effort into their generations (iterative, manual cleanup, etc) figured out the extra fingers thing ages ago, but people using Bing AI or Dall-E without knowing how to properly prompt will probably still get them.
Yeah that's not really true anymore, anyone who knows what they're doing can very easily mitigate that problem to the point it's a non issue. 🤷♂
few months ago this would be true
Another thing ai does a lot, at least right now, is that it feels like there's a color filter on everything. Everything is desaturated such that it looks like someone put clear paper over it. If you take these images and manually turn up the saturation it looks way better and like that's how it was supposed to look.
I’m not illustrator but really appreciate the hard working artist. AI kill talented artists
Great video Veno. 16:59 This plagues not just drawings but art in general, and AI 'Art' is the ultimate version of that.