#art discourse

[funishment-time]

"art is for the TALENTED only TALENTED PEOPLE could do art before and AI artists are HEROES" why do all AI Art Bros sound like Nagito Komaeda

[dominaexmachina]

“Click to get a result”: Who told you that?

Everyone yelling “AI art isn’t real art!” seems to believe they’ve cracked some universal truth — that anything made with a tool must be effortless and brainless.

But here’s my question: who actually told you that prompt + click = result?

Was it Midjourney’s Discord thumbnail at 512x512?

Was it some meme of a soulless cyborg churning out anime faces?

Or was it your own lack of experience masquerading as a hot take?

Because let me walk you through a basic “click to result” StableDiffusion flow:

First, you invent a scene. You visualize the lighting, expression, tension, texture, story — a full-blown internal composition.

Then, you write the prompt. Oh, and not just one. You write 30. And you tune them. And you test what each word actually does.

You get a “draft”? Congrats. Now you start fixing: Inpainting the face that melted like a crayon. Outpainting the background that doesn’t match your story. Rebuilding the pose because the character has 3 elbows and no spine.

Then upscale? Oops, it ruined the eyes. Fix again.

Then composition shifts? Fix again.

Then color grading? Details? Harmony? Fix. Again. And again.

Until the scene feels right — not perfect, but alive.

Now tell me… what part of this was “just clicking”?

This isn’t about “soulless machines.”

This is about people who think effort = brushstrokes and nothing else.

People who think their suffering is what makes their art real.

And people who refuse to acknowledge that creativity can live beyond the tools they were taught in college.

❝Creativity is not a medium. It’s not a job title. It’s not your suffering. It’s the ability to turn a vague feeling into something you can share — and that doesn’t go away just because you used a model.❞

So tell me again —

What exactly makes your process real, and mine not?

[curioscurio]

how yall look standing next to million dollar paintings in museums

[vijigenshin]

Attention JTTW, LMK, general fans of Sunwukong, @xi4oyan is using AI for their 'art' posts.

I love being in fandoms and scrolling through post only to be uppercutted by people pretending to be artists!

Important Note: this only involves their 'drawn' art pieces and NOT their writing. That's not to say we shouldn't be suspicious of the validity of their writings. I just am not equipped with the knowledge to catch AI written and or AI edited writing.

Ok warning over im not beating around the bush. Since this person clearly has a bit of a following AND they do a half decent job of covering their ass about this so im going to drop the nuclear example first.

Theres a lot to unpack here but I'll start with the 'shaky' evidence by saying this is an AI art style. I made a post a month-ish ago about a person named @scrollsbymoonlight who also uses AI to post 'art' they made. This AI's 'style', specifically how they render, is AGRESSIVELY similar to whatever AI Xiaoyan ( xi4oyan's name according to them) uses.

This is an example of 'art' created by whatever AI Moonlight uses as an example. I was able to clock both pieces immediately as AI from the style ALONE, since this weird anime, almost studio ghibli, watercolor style is one popularly used.

Now im sure you noticed it already but I circled and color coded suspicious or downright incriminating aspects of the 'art' shown. Here is my elaboration based on color.

Red: this weird faded out of place line is the AI fighting with itself in regards to whether it should place a collarbone in that spot, since often times a collarbone is indicated in this spot, but obviously even the AI knew that would be wack ass placement which is why its weirdly faded and just wrongfully placed in general.

Orange: inconsistent line work on her face stripes. One is tipped sharp, one is weirdly rounded. AI is terrible with pattern consistency and recognition, and no human would make this kind of mistake, especially at this skill level.

Yellow: the background has many a melting sunflower and sunflower leaves. AI is also terrible at background clarity, which is why the more faded sunflowers look so melted and warped. Oh also the sun is warped on its right side on top of being barely visible, something an actual artist would never do.

Green: inconsistent shapes for the stripe pattern AGAIN, AND the bottom stripe cant decide if its one stripe or two.

Teal: this one is lowkey kinda funny, there is a VERY faint outline of what looks like to be another finger here. I feel I don't need to further elaborate.

Blue: the true ai classic,,, ✨️hair melting✨️the hair strands are hard for an AI to deal with, so often times shit just melts together or overlaps in unnatural or odd ways. The lower blue circle has another less visable connection line for the hair that is extremely out of place, and the top circle is just a mess of connected and disconnected lines.

Magenta: this area is the most obvious evidence for this trainreck! The hair is melting into her kimono top, the AI wants to put a second sun here but it never fully comes through and the circle is unfinished, the circle I question confuses the AI and thus has hair follow the curve of the sun in a way no human would draw, and finally the split in the hair here follows where the line of the sun would have gone had the AI finished it.

So as stated before, Xiaoyan likes to try and be subtle with their posts. They choice pieces that are hard to catch, and later on you'll see a piece they traced over with fake proof to further give off the illusion of validity. Sketch AI generated images are hard even for artist myself to prove, since the entire point if a sketch to to not fully care about errors.

However that does not mean they're immune to flaws from the AI's end, as well as AI style.

Aside from the fact that this is a blatantly different artstyle from the one previously shown, there's a bunch of curved and sketch lines that no person who is actually sketching would ever actually make. They're near invisible and you have to zoom in and play spot the difference to notice them, and even then if you're not a seasoned artist you still wouldn't catch it.

Cleaned up sketches struggle to look so clean, and even then this looks like its trying to emulate non digital art and I can guarantee you if that's what its supposed to be then its even more suspicious since tradition pencil drawings are SO MUCH MORE MESSY.

This artsyle just also screams AI. Im not counting that as fool proof because even though I have proof for other things saying 'It LOOKS like AI trust be bro' feels disigenuine even if its very true. Im only saying it here because I still think it should be mentioned.

Now onto our regularly scheduled color coding!

Red: y'know the more of these AI pieces I study, the more I realize that AI just doesn't understand the collarbone... it's trying to be a contractor muscle that doesn't need to be there and you can TELL because THERES TWO RANDOM ASS LINES COMING OFF OF THE UPPER LEFT COLLARBONE LINE.

Orange: that fly is not properly connected to anything. AI doesn't fully understand that clothing folds are not clothe break ups.

Yellow: the lines for his knuckles are melting in all sorts of wrong way. Also remember those hard to spot random squiggly sketch lines? There's a few of them here.

Green: a weirdly clearly drawn line that has an uncesecary split. Seriously there's no fly away sketch lines that would indicate a human drew this.

Teal: AI does not know how to deal with men who have nice tits 😔 fr though his pecs unaline from the moment the shirt starts and it curves in WAY too soon. Likely because AI doesn't like overlapping its more recognizable lines, so it compensates by flipping anatomy the bird sometimes.

Blue: melting hair that cant decide if its hair or muscle definition, muscle definition that doesn't exist in this spot so...

Bonus!: on the left side of WuKong, from his eyebrow to the first 4 or 6 defined hair tufts there's a lot of melting. Sketch drawings have some cute hatching sometimes on the outside of the main sketch focus, and while the AI can understand most of it, it has a hard time not turning hair strands into said surrounding lines

Important note: due to Tumblr image limit i cant for the next images shown, I cant post the proof that Xiaoyan posted the pic in this post. I do have the images if necessary for later though.

Man don't you guys love incredibly random art style changes?

"But Nijira" I hear you saying "Sun Wukong has been depicted multiple times in many different body types and facial structures! Isn't it possible that these sketches depicted two different Sun Wukongs?"

No its not. This person tagged both images under only a few very popular depictions, all of which are wildly different. This means this is just a general drawing of the character, and not a concrete version, thus the abrupt style change makes no sense.

"But Nijira!" I hear you call "isn't it possible that this is how they used to draw Wukong? Couldnt they have just changed their preference?"

Maybe if it was over a long period of time but no, 3 days later Xiaoyan posted this:

Which is so aggressively different their style used 3 days ago that this is not possible.

"Nijira!" I hear you exclaim one last time "are you going to be deep diving the piece above?"

Not not that piece, simply out of love for my fingers as I have the previous image to describe alongside two others. To be clear though, the example image is likely traced over AI and not fully AI itself, but unless the people ask for a follow up to this post, I will not be breaking that statement down, meaning you're free to take that claim lightly.

And now back to our regularly scheduled color coding.

Red: his kimono's seam randomly disappears.

Orange: really bad melting on the line art of the tail.

Yellow: this finger is fucked five ways from Sunday. Alongside a random not in style finger fold, there's also a random crack(?) On the inside of his fingernail. Wukong should get that checked out...

Green: two lines are trying to be the outline of monkey's tail. Also there's some melting here.

Teal: random line coming off of Wukong's bracelette, and 3 random lines outside it.

Blue: the finger linear is melting into his face lineart. Also the AI tries to give Wukong's eye another pupil in the corner.

Magenta: bad melting and random line that has likely the AI trying to give Wukong's kimono a shoulder seam.

Holy art style changes batman!

This image is hard to breakdown because I cant tell if its straight up AI or if its AI that was traced over, something Xiaoyan has done before and will be shown in the next image.

Theres also a possibility that this was AI that got edited. I have no proof for this outside of how clean the greenery around her is.

theres another possibility that this is a drawing 'enhanced' by AI. Some programs allow you to feed art into it to allow the AI to 'improve' the work. I don't think that's the case but its always smart to consider all options.

Go into the color coding proofs and make your assumptions from there.

Red: the obi (sash) of her kimono is also apart of the kimono's skirt(? Is it a skirt? I think its just extra fabric cut off by the shirt? I have no idea if this has a technical term...)

Orange: random ass dot on her eyelid.

Yellow: this part of the tiger cant decide if it wants to be skin folds or stripes.

Green: tigers mouth is fucked up so you cant tell wear the teeth linear starts and ends.

Teal: This tiger now has an ingrown nail! Seriously though, the rest of its claws protrudes, this one doesn't likely on the account of the whole AI doesn't like intercepting lines thing

Blue: the nose linear on the left side is weirdly think? I THINK the AI is trying to indicate the other nostril, but at this angle you wouldn't see it, something someone who can actually draw at this skill level would know.

Megenta: ignore this, there's nothing wrong here. I thought it looked off but on second inspection it doesnt.

We're finally at the end here! I can almost stop typing! You'll notice there's no indication lines here and that's for one reason; this image ISNT about the AI used, its about disproving the 'proof', specifically the guidelines, this post tries to show.

And can I just say how SAD this post is? Because aside from the fact that this is traced, the rendering for this isn't bad and its definitely not fully rendered by AI. Xiaoyan if you're reading this, you HAVE POTENTIAL. YOU CAN BE AN ARTIST. You don't HAVE to use AI to feel good about your ideas. It might not be at the level you want it to be, but it'll still be better than this souless AI slop.

This is not a hit piece on you as a person, its on the fact that you use a medium that actively exploits artist to pretend to be one.

You CAN do better, you just have to be WILLING to.

Of course my job isn't done until the explanation is written.

I've mentioned before that I myself am and artist, I sketch, I know how it goes how it works, and let me tell you- this is the WORST attempt at fake guidelines I've EVER seen.

Nothing connects like it actually should for someone who would be at this level. Their sketch draws the shoulders as if it's connecting the kimono already, when the entire point of doing guideline sketches is to take care of the anatomy BEFORE LITERALLY ANYTHING ELSE.

The guideline for the direction the face is tilted doesn't add up properly, the right elbow joint indicator is the same width as the sleeve and not the art it should be indicating underneath, the spinal curvature indicator doesn't match the posture, the line indicating the chest counts for the kimonos, the upper circle indicator is useless.

AND THAT'S NOT COUNTING THE FACT THAT THE NEXT 'STEP' OF THE SKETCH IS JUST THE ENTIRE LINEART.

No anatomy drawn before hand, its barely messy, at this stage the line indicators are only still present to make it look convincing- this is so obviously fake its PAINFUL.

So yeah. @xi4oyan uses AI to pass theirs elf as an artist.

One last message to those who liked, reblogged, or even followed this person; please don't feel bad about doing so. I understand being frustrated about not being able to tell, but its clear this person jumped through hoops to make that convincing. You are at no fault.

[joyce-stick]

Think it's past time we started calling the "yuri is fetishizing lesbians for cishet men" argument what it is: misogyny. It's misogynist to say that women can't genuinely create or enjoy lesbian fiction

And it's also teetering on the edge of transmisogynist because that argument is like two hairs' breadths removed from "and therefore any trans women who say they like yuri must actually be men"

(also even if it's true and men like looking at lesbians: so what. Who cares. They're allowed to like it)

Just admit that you're homophobic and think porn is evil or whatever

[firecooking]

STOP IT STOP IT PLEASE STOP COVERING UP OLD ART AND DESTROYING YOUR OLD SKETCHBOOKS AND 'FIXING' THEM PLEASE TIK TOK IS WRONG YOU WILL MISS THAT ART SOME DAY AND YOU'LL BE MAD THAT SOMEONE CONVINCED YOU TO HATE YOUR PROGRESS

[the-crooked-library]

look i don't know how true this is but i hate hate hate in my bones the way that the recent rise of purity culture has intersected with minimalism and irony poisoning

no sex. no dark themes. no "fetishization." no ornamentation. no "unrealistic" lighting and you can't see a thing. actors don't enunciate for shit. no "purple prose" or "overwriting" (it's just an adverb) or "tumblr prose" or whatever the fuck the newest term will be. sincerity is cringe, ew, why would he say that. no colour grading, because we want the big-screen blockbuster to be "realistic," BUT we will shoot it in a room with fake sunlight and slap the effects onto the scene post-production. moral fiction. moral fanfiction. "omg this is craazyyy was the creator on drugs??!!1????" about any form of creative expression. lists of reasons why this short experimental amateur one-shot is Very Bad Writing, actually. s*x smex spice adult fun time p0rn k!ll grape sewer slide.

everything is boring, nothing is real, i am fucking sick of it

[garlicbreadslice]

“I use gen ai because I want to make art but I’m not good enough at it to make the art I want :(” well friend guess what. You’re still not making art. And you’re still not good at it. So like. What did we learn here.

[xnixnbvg]

I know we all hate AI but here's a poll I have designed to see what artists and others think, tracing and ai art are the most controversial art drama there is. Now here's the question...

Keep reblogs and comments respectful if this goes viral.

[leathercircuits]

when i was 10 i used to think all artists would immediately be best friends due to their shared interest. BOY WAS I SO WRONG 🤣

[moxalotl0]

No matter how bad your art is, at least it’s not ai

[dominaexmachina]

Austin Kleon “Steal Like an Artist” vs AI Training: Same Game, Different Players

Everyone in the creative world has at least heard of the book “Steal Like an Artist”.

Even if they haven’t read it, the title alone became a viral mantra — bold, catchy, and unapologetically rebellious.

Published in 2012 by Austin Kleon, this small but punchy book quickly carved out its place on the desks of artists, designers, writers, and idea-makers everywhere. It’s been quoted in classrooms, conferences, and countless social media posts that say: “You don’t need to be a genius. Just start copying — then make it yours.”

And now, as people scream that AI is “stealing,” let’s take a moment to revisit this beloved creative classic.

I’ll quote some of its key ideas — and show you what AI models are actually doing.

Let’s see if it’s really theft… or just familiar behavior wearing a different skin.

“Nothing is original.”

Models don’t generate from scratch. They remix patterns from what they’ve seen.

“You are a mashup of what you let into your life.”

AI is a mashup of the training data it was fed.

“Collect good ideas. The more good ideas you collect, the more you can choose from to be inspired.”

AI collects patterns from a huge dataset. It doesn’t know what’s “good” — but it maps what people liked.

“Don’t plagiarize, transform.”

AI doesn’t copy exact works. It creates new outputs based on statistical association.

“You’re only going to be as good as the stuff you surround yourself with.”

AI’s quality reflects its training data. Garbage in, garbage out. Just like with people.

“Draw the art you want to see.”

Prompt the AI with your vision, and it draws the art you imagine.

“Creativity is subtraction.”

Fine-tuning a model or crafting a prompt is subtracting until only the idea remains.

So what’s the real difference?

Humans say “I was inspired.”

AI says nothing.

And that silence? Oh, it burns some people — because it exposes how much of human art is remix, too.

You love the idea of genius.

But you hate it when a machine mirrors your own methods — without ego, pain, or your romanticized suffering.

[deepdreamnights]

Hey there, saw your post re: harassment around artists using gen ai and thought it was great esp with the debunking of data usage myths. Would you share your thoughts regarding concerns that models are being trained to copy specific art styles and thus pose a direct threat to the artists whose art styles are being used?

Well, there's several levels to that.

The main one is that on copyright grounds, styles are explicitly non-copyrightable. Moreover:

No one's style is unique

No one's style is unimitatable by analogue means.

The second point is important, because anyone can go on Fiverr right now and and find someone to replicate any given art style, and every competent draftsperson has to be able to do it to some degree or another. No major animation house, art studio, or comic company has ever hired someone because they couldn't find someone else that could imitate the surface-level aspects of their style.

The first point is just a matter of basic reality. Ex-nihlo creativity either doesn't exist or is so rare as to be a once-in-an-epoch thing. Everyone builds on the influences that they learn from, and if you think someone has a unique style what they really have is a different media diet than you.

For example, Don Bluth. Born 1937, aged 15 in 1952.

Same year Time released this this picture of Burlesque Performer Dale Strong.

Someone made an impression.

Marilyn Monroe was also a national sex symbol when Bluth was a teen, putting some context to most of his other ladies, but especially Goldie Pheasant (or maybe she's more Jayne Mansfield, hard to tell through the bird-ness). His art style has obvious roots with Tex Avery and I would guess he read Mad Magazine a lot as a kid.

And Not to hang the guy out to dry alone, I was a teenager in the 1990s, and most of my sexy fictional ladies are 9/10 some combination of Dana Scully, Peg Bundy, and Rhonda Shear.

The point being that style isn't something you create intentionally so much as an accumulation of influences, drawn from the commons. Attempting to claim ownership of such a thing is by itself an act of theft in my view, and allowing them to be protected under the law would mean a judge being shown exactly how many pieces of prior art the Walt Disney Corporation owns that your work superficially resembles. Why, they'll even run it through a style recognizing AI to make sure they catch them all.

But let's talk about style matching.

It just takes one image now, and doesn't require training.

Which I'm sure sounds frightening, but this has been the situation since February for Midjourney, and it was available in the Stable Diffusion ecosystem long before that. If the threat were as pronounced as feared, we'd have seen the impact by now. And we haven't, and we're unlikely to, for several reasons, several of them listed above.

The largest is that style isn't even close to the be all/end all of what an artist brings to a given project. And the kinds of execs who are making a 'replace 'em with a robot' kinda decision aren't the kinds of people who care about art style beyond how much it looks like the most recent successful thing. And nobody's ever needed a robot to ride coattails.

But the next largest part is that AI style imitations aren't really accurate because the robot doesn't see style in the same way we do. It's all just math to the robot, and it prioritizes what it notices, not what we do.

I'll demonstrate.

Jack Kirby will be my example, for several reasons.

He has a bold and identifiable style, he's arguably the most famous artist in western comics history, and he has many analogue imitators and homagers.

Using Midjourney and prompting "an illustration of dana scully by jack kirby, 1968, in the style of 1960s marvel comics --ar 3:4 --s 15"

Using the base model, on the first roll we get three complete style mismatches and one that's kinda close, though I'd say that's way more Sal Buscema or John Byrne.

Kirby's women had a certain, difficult to describe oddness about their faces that the robot doesn't seem to grok, and it doesn't touch on the kinds of wild patterns and bold black/white swatches that make Jack's work feel 'jack'.

Tom Scioli's take on Kirby is a sort of lovingly flanderized parody, but it captures the spirit of Jack's art much more directly even if a lot of individual details aren't period-accurate. He draws Kirby the way you remember Kirby from your childhood, but I don't question whether the page above is trying to be a Jack Kirby homage or one to Sal Buscema.

But Midjourney has style reference, so we can inject the Kirby right in. Using the picture of Sersei dancing from above with the same prompt, we get:

Well, the work is more convincingly period, but again, we're not terribly close to being on-point. In fact, they're not very consistent between each other. Top left is any 80s marvel fill-in artist. Top right is maybe Kirby-esq. Bottom Left is flat out Jim Lee, bottom right is very Byrne-y.

Using three reference images to give the best shot, I'm also moving to using images of a similar color style, and all with a woman as the central focus. I have included the infamous Crystal pin-up shot because as I said, Kirby women have a certain oddness to them (fondly).

Results (MJ 6.1 on the left, Niji 6 on the right):

It all says 60s-70s Marvel, but I don't think Kirby would be the first guess for any of them. Maaaaaaybe the lower-left Dana in image #2 if you squint.

And that's Jack Kirby. Massively popular and prolific with a career spanning decades. If anyone in the comics space should be impersonatable by this thing, its him.

I'm sure you could train a LORA to get closer, and sure, the tech is only going to get better from here, but by the nature of how the system works no generation pulls just from what is referenced. Every generation is both blended with other concepts and emphasizes only what the machine catalogs as relevant, not what we might.

There's not much to stop someone from imitating your style with a machine, but there was nothing stopping them from doing the same with an underpaid freelancer. The results are likely to miss the mark regardless.

If the client wants you, they'll try and get you. If they just want something kinda like you, they've always had an avenue to that.

Fortunately, you're more than your style, and whatever anyone can do with the machine, you can do better because you've got access to both.

[mordenandmerry]

Some people draw skin so pale I’m wondering if they’ve looked at a person in a while

[lillikoifish]

Theres a lot of discourse on modern art, and its cool to see the conversations happen, but sometimes people get so lost in the sauce that it actually crosses over into just kind of depressing. There’s this piece by Yves Klein making the rounds on tiktok, being thoroughly roasted by 15 yr olds because “It’s just blue on canvas!! That’s not art I could do that!” And there’s tons of people explaining the lore behind the work, he invented that shade of blue, if you think you can do it you gotta invent a color, which no way you can do, liar.

But that to me isn’t the part that really sticks. Honestly let’s take all the lore out, just remove the art from all context and just look at the piece in question.

How long did you look at this art before you read this sentence? A minute? A couple seconds? That’s the art. That’s the real interesting thing. At the end of the day it’s just a shade of blue, and if you like you can enjoy this simple pleasure: you can take in a gorgeous color for as long as you like. How long did you get to let your eyes rest on something pretty, just for the sake of it? Maybe you wanna look again now, because it’s been a long day, you’ve earned some blue.

What these kids fail to realize is that they are denying themselves that. Are you so void of all humanity that you can’t just enjoy the color blue? You need to be impressed by skill, the art has to prove itself worthy, it has to fulfill a need, it can’t just make you happy with a pretty color, and if you can do it yourself, that just speaks to how pointless it is, because anything you yourself make cant have any value, right?

This kind of self hatred is exactly why works like these belong in museums, why they should be in front of your eyes, because a person doesn’t need to have value to exist, you can find whatever joy in this life you like and it doesn’t have to be for any reason at all. You can just be on earth and enjoy your time here.

And hey, maybe you just like the color blue, here, why don’t you look at this canvas a while. I think it’s very pretty. Don’t you?