#Ai art critique

[sannyo-appreciation-posts]

Sannyo Day Special

Spot The Sannyo,

Sannyo vs The Machine

A game of Spot The Difference.

(More just a game of spot A.I being stupid)

(If you just want to play the game without explanation, just skip to the line)

A.I has become a bit of a plague, so its nice to know characters like Sannyo, and as another example, Enoko, will give it a little bit of an existential crisis. Leading A.I editors to need to throw away many attempts or cut them up.

There are a surprising number of unique things Sannyo has that confuses ai, so keep an eye out for them as you see them.

All my examples can already be found in the Google image search category. I've also picked mostly, what appears to be unedited examples for the maximum number of "spot the difference" points

____________________________________

____________________________________

Here's our game of spot The difference:

the rules are pretty simple. Anything that an an A.I can be spotted screwing up, they can be called out on,

but if it something that could pass for a stylistic decision, we have to ignore it. As simple examples,

Everyone draws her clothes pattern kinda different, so we can't be picky about that.

Sannyo wears multiple layers on her primary outfit, but some artists interpret these as a hem or fold instead of an additional layer of clothing. So that is acceptable

(The solutions will be immediately after the image so scroll slowly)

[Sannyo examples, Zun art and Dari art]

Here we go!

#1

_______________________________________

#2

_______________________________________

#3

_______________________________________

#4

_______________________________________

#5

_______________________________________

#6

This is not Sannyo's outfit? Bot failed to complete the assignment. Get out of my classroom!

_______________________________________

#7

_______________________________________

#8

_______________________________________

#9

_______________________________________

#10

_______________________________________

_______________________________________

So I've compiled a few things I think confuse it the most:

If you spot any more I missed, comment or repost it with it. I'll eventually edit more in.

[everythingisok3000]

Dude I prefer anything created by a child in MS paint fifteen years ago to literally anything created by AI. When I look up Cyborg Pigs I'd rather see this:

every day of the damning week than ninety-percent of what's being shown to me on google images. There's just literally no artistic vision there whatsoever. Where's the desperation? The drive to make something with your own hands? You can't MAKE me like AI art man.

[titleknown]

Gotta say, while I love @mostlysignssomeportents' work, I think I gotta hard disagree with their sentiment that AI art comes "pre-corporatized/gentrified".

Like, that might be true if one treats the one-off images one sees as all you can do with AI art, but from what I've noticed in the people doing what I consider good work with AI art, from @teuthisdreams' work with collage on his most notable comic and general openness about his process, to the layering Shenanigans that @reachartwork does to avoid the Standard AI Art Artstyle, to @therobotmonster's use of good ol fashioned hand-done retouching and photomanipulation and advice on such, the singular image is not the primary artistic use-case for people serious about the tool.

Rather it is that AI art becomes art in its relational capacity, combining with other tools in the nature of Eisensteinian montage and collage, the latter being a bit ironic given the fact that AI on its own isn't really a collage machine like people think it is.

The same way that, say, some stock art for Apple stuff could become the basis for Jerkcity or a piece of stock art of a generic 50s-type guy became the notorious mock-religious-messiah JR "Bob" Dobbs, or how standardized sliders in a video game made the basis for The Final Pam or the Boy-Mayor of Second Life.

Even if the outputs are to be considered generic, the way they are used to relate to other things by the artist, and the way they relate to other things the artist has done, is the way art is found. The humanity is added in the editing and juxtaposition, especially if you're combining it with other mediums.

And I think the way we're trained to treat art as isolated; decontextualized snippets in an algorythmized feed is the actual problem that I think the fears around AI art as an artistic practice are around.

Now, there's a lot to be said about that, from the issue of how the way social media discourages archival/archive-binging is a far greater contributor to the depersonalization of art, to the ways in which algorithms hide artists who can't keep up with them and that needs to be a labor issue we push against.

I have a longer piece on the nature of this that I've got bouncing around in my head that I need to finish after Artfight and Kaijune-in-August at any rate.

But, I think I'll end with the fact that if you're looking for the grungy endearing grassroots jank that Doctorow feels is lacking in the uses of these tools, I think that the best place to look first would be at stuff like this:

[thoughtportal]

[foxgirltail]

Palworld critiques are looking more and more like Tumblr call-out posts they're all like

Long, drawn out, effective non-issue with several sources/instances

Another non-issue

This one is actually of substance and should be the main point, but it's buried by minor issues of little consequence

Another non-issue

[mauesartetc]

Redrawing Shadiversity's AI Piece

For context, check out this post here. This is, uh... It's a doozy.

Let's start with the main character of the image. The girl's pose looks very awkward and unrealistic for what she's doing. Her feet are dragging in two different directions that don't indicate the direction she's jumping in, and it looks like her top half is getting blown back in a wind tunnel. According to one of the reblogs on the post that introduced me to this thing, the pose wasn't the generator's doing, but the artist's. "He drew the girl and photoshopped in a picture of a lizard and a picture of a church and had the image generator "refine" it."

I sincerely doubt he used any kind of photo reference for this drawing, as it'd be uncomfortable for anyone's spine to curve backward like that while they're leaping forward and swinging a heavy sword. That just looks painful.

Let's explore some ways we could make the pose look more believable.

I think I'll go with a pose that's close to the original but makes a bit more sense.

It obviously doesn't have the same level of... "polish" the AI version does (we'll get to that in a minute), but the tilt of the spine looks much more natural for the direction she's leaping in and the way she's holding the sword.

Now that we have that out of the way, let's analyze more of the image as a whole.

AI art handles detail in a way that looks good to the untrained eye, but falls apart in the eyes of experienced artists. These clothing folds, for example. There's no logic to the way they're shaped, and the shirt is randomly tight around the chest when it's loose everywhere else. Then there are the scales brought into sharp focus despite the rest of the dragon being blurred, the blood drips that look like stalactites, and so on and so forth. I'm sure there are things I missed, as well. If y'all find them, let me know in the comments!

Something to note about the sketches I made before the finished drawing: They kinda suck. And that's the point. The early stages of a drawing aren't meant to look pristine with perfect anatomy (not to say the finished product is anywhere near perfect, but still). What they are meant to have is energy. Purpose. Life. But AI bros are so afraid to make any "bad" drawings that they don't draw at all (or in cases like Shad's, they only draw the bare minimum).

I didn't make this post to dunk on AI prompters, but to encourage them to put in the necessary work that will improve their skills. And no, I'm sorry, typing words into a box won't make anyone a better artist. It might make them better at describing what they want when they commission an artist, but by and large it's like lifting a feather when you want to gain muscle instead of, y'know, lifting actual weights.

Obviously machine learning isn't going anywhere and it'd be nice to use as a tool to make different steps of the art process more efficient. It's good for silly memes, I guess. But we shouldn't treat the images it spits out as masterpieces, and, importantly, businesses shouldn't use it to replace real people.

Anyway, it's pretty easy to go to the store with five bucks and come back with a decent sketchpad and pens/pencils. Not to mention art programs like Krita and Blender are FREE, and there are plenty of tutorials on Youtube. Just sayin'.

Get drawing.

[furiousfinnstan]

I saw somebody made a whole short movie with mostly ai and then I found this again...so decided to finally publicly post it tho idk how I feel anymore

[venndaai]

I'm not done with Tacoma yet but it is really good and I'm mad at myself for not playing this six years ago when it came out.

[illyth]

Something I drew during the drawfee stream earlier today that I'm actually really proud of. I have basically no clue what I'm doing and I am sure there are a trillion mistakes, but I really wanted to start learning how to make art at the start of the year, and I've been in a rut with it over the last month. But I felt inspired tonight to make something and I really like how it came out!

Inspired by the character Hikari from Octopath Traveler II and made in clip studio

[arcanesouls]

I run a space for creators, coworking, and learning. Challenges every month. Critique, portfolio building and more. Join us!

[firedragon1321]

[22degreehalo]

One thing that I always go back to when I get to thinking about machine-generated art

When I was in primary school, a friend of mine was really good at art. Like, not 'can carefully copy the Neopets Uni artwork' good, actual 'winning awards for her art piece' level good. (Look, for a horse girl back then I think that was not nothing of me!!)

And one time, when we were in year 5 or so - meaning the early 2000s - she entered an art competition with a beautiful picture of a bunch of African animals eating at a table together with the sort of cartoony, whimsical style you could easily see in a published kids' picture book (or so I remember it, anyway).

But during one of our classes, a guy came to the door to speak to her. And then she came back into the class to her cubby, awkwardly and embarrassedly carrying her picture.

Why? Because she's coloured it digitally, which (unawares to her) was against the rules.

She'd carefully sketched out the piece and drawn the linework physically, on A3 paper. And everything - the composition, the art style, every little detail - was her decision.

But she'd used a computer instead of paints or pastels or colouring pencils or whatever. And to the runners of the competition, that was 'cheating.' You weren't doing it yourself. All you had to do was click a few buttons and the computer filled it in for you, automatically.

Even at the time, that seemed wrong to me, though maybe that was just my inability to believe that my artist friend could've ever been wrong about art. But as I got older and tried out my own digital artworks and colouring... yeahhh. It's a hell of a lot less easy than I thought it'd be, if you want to make something look like how you can imagine in your head. To this day I think that my decades of doodling and anime art have made me okay at lineworks, but I never learned shit about colouring and it shows.

And it's just. I don't want to perpetuate the mistakes of the past, you know? Learn from history. The new technology is always going to seem scarily 'too perfect' and 'too easy' and unskilful and dangerous because What About All The Real Artists Who've Struggled Hard To Perfect Their Work? But digital colouring didn't kill traditional methods. They're just different.

(and requisite statement that I'm not thrilled about artists losing jobs but also by and large these would be the dead-end generic boring bullshit jobs nobody would be excited to do anyway soooo this is 99% a jobs and economic issue that has already happened before with countless past technologies and there is absolutely nothing to gain from hyperfocusing in on ai instead of, like, UBI or something to actually fix it.)

[titleknown]

Given my defenses of AI artists/dim view of the moral panic, I think overall people pushing back on this is a good thing.

Like, the sort of AI art I like is by small weirdos doing weirdo things not big megacorps using it to replace workers, and WotC and Hasbro are run by bastards who would probably salivate at the idea of replacing artists on MtG given they'd probably be the easiest ones to replace from a greedy-corporate-bastard view.

However, there's a souple of notes here I think we all should keep an eye on.

Namely, the fact that it was apparently from an artist using an AI tool that WotC didn't know about. Which, I think that's going to be a problem creators run into sooner or later with "AI bans"

Like, I get a lot of images for my photomanip stuff from Pixabay. And more and more I've been seeing a lot of AI art stuff as a part of it. How would those bans impact me if I; say; used them as an ingredient in a photomanip without knowing?

So, there's that. But there's also a disquieting possibility I've noticed.

Namely: Given the main fear about AI is that its ability to work quickly at volume might be used to push out traditional artists and their precision, what happens if corpos push for "the worst of both worlds"?

IE, what if; due to the expanded production speed of AI art; they mandate that creators work at speeds that can only be done if they use AI tools, but also end up forcing them to conceal and lie about it and leaving the artists to take the fall for using it rather than the corpos for pushing the culture of overwork that made it necessary?

TBH, I think that's why we need to focus less on the "stealing" argument being used as a way to talk about the destruction of the artistic ecosystem by mandated speed, and more the idea of the right of the artist to work at their own pace that might be able to do something about it.

But, IDK, that's just me, it's just something to watch out for.

[ultimate-saulo]

So I tested Glaze and it's a wonderful new tool to protect human art from AI scrappers. But the "invisible to the human eye" is not very good with flat colors or cell shading art styles. I wish I did very rendered or textured art so the artefacts that it makes doesn't show too much but welp, my style and the style that better fits merch is more of a cell shade flat colors kind of way. I will definetly use it in FULL rendered illustrations but other than that I hope it gets better "invisible to the human eye" in the flat color style.

(Left is very low intesity and right is low intensity. They recommend better protection with low intensity but the image gets more altered. I hope very low is better than nothing yet still it shows the weird texture.)

Here's the original for comparison:)

I really hope the software gets better and it can achieve that "invisible to the human eye" promise. And I'm SURE it's really really very good with art styles more textured and rendered to the gods. I just wish it was better with simple art styles in the beginner level.

EDIT: It's actually really good on phone screens I hate different screens in different devices 😭 Might actually gonna use Glaze in cell shading work if it ain't laptop-screen-bad 🤞

[greenpanda-officialart]

Ergojosh :Use of Ai Art

Kay, I don't usually make posts like this, but its on my mind so here we go;

Ergojosh, an art youtuber whom I've followed created a video using ai in his art piece. He created his own original work, and only used ai as a tool- this post is nuetral in my opinion. I don't agree with Ergojosh, but I don't think what he did deserve to be cancelled and condemened, it's something a lot of artists had thought about (I'll get into that later)

His first initial sketch- with using a veil for the face and overall it looks like his art style, Ergojosh has his own unique signature that makes you think thats HIS art.

Then after using Ai as a reference the face changes, looking at this I would assume scott campbell made this, already you lose his identity by using other references and yes the sketch is looking better visually, but you no longer see Josh in the artwork.

The worst part at 3:58 he doubles down and becomes defensive that he used ai as a reference that he didn't generate it-

Which then later on when he admits to using Midjourney using a program and putting money into it and endorsing said product kinda defeats the purpose? "Oh I didn't generate ai therefore it's okay, anyway I generated an ai on a paid program that profits of stolen artworks to make this generated image." that artists are not okay with, so why be defensive of using ai refs when he generated images after? Oh, because he said it wasn't the 'face' he removed the faces but used it for hair, be it hair, eyes or anything a generated ai image is a generated ai image. No matter what it looks like.

So this is the final pose, which looks okay, but the hands should've been holding the hair strands and more hands couldn't been around the body in a sphere composition holding the strands.

around the 10;00 mark in the video, he sounded aggresive about the use of ai, that it's a tool, we should adapt and be mature. But people are allowed to hate ai and not want anything to do with it and would be upset at people who endorse it, after all if a small artist can use a few seconds of a song in their youtube videos they get copyright striked.

Meanwhile artwork is just scrapped and used- its rules for thee but no rules for me- type attidude, that the program and companies would not want their work to be stolen but think it's fair to take others works because it was public.

Even so- I get why Ergojosh was only making a video of using ai as a tool in his creative process, that yes the hair was a challenge, it's like get a glass ball and put hair inside it how would that look- oh with how broken the internet is thats a hard to find ref that might not even exist- the problem though.

Is that it's clear Ergojosh was trying to force ai to work when it wasn't happening, the truth is I think as an artist you can't make ai work for you- when I was using it for Uni, I felt frustrated at the lack of control and that it wasn't working out, people who aren't artists would be fine with it, but as creators its just more work.

And it was obvious in this video that it felt as if he was making it harder for himself.

so first image is the ai generated image from midjourney that he used,

second is the traced parts. I get it, there are some shapes in the ai image that looks good but this image just looks messy, the hair doesn't fit or flow like it should there are some strands that are too detailed/clutered and some too simplified.

after using grids and getting the composition down, he then fleshed out the hair and well;

This honestly felt like a cry for help, how he keep saying "I am the producer, I'm in control-" this hair looks random and messy like, was there any need for an ai input at all? Any one could've done this. there's no direction, no flow just random wisps just slapped on there.

He changed the character to a gremlin with a tail, worked on the lighting, but at minute 20 I was thinking...this is a lot of headache what a convulated way of working harder not smarter, it felt like an obvious struggle and the wispy tail was lost in the hair esp with how thin it was, and IF the tail had the hair draped over it instead but it's not its just, floating there.

And heres the final image which uh-

I don't understand how someone could spend possibly hours on an art piece that looks like ai-

The hands above holding the hair strands are gone, the eyes look like ai, the hand holding her chin also looks ai generated and at a glance nothing about this artpiece screams ergojosh.

some hair strands were coloured but some like the face look like a diff artstyle and the strands surrounding her look unfinished-

The only thing saving the focal point is the lighting, but even the two diff coloured eyes?

It felt like a lot of effort for something that didn't need the ai assist. Look lots of artists have differing opinions. Most hate ai and all it's uses even as a reference it shouldn't be entertained to even be called that. Most try to steal Ai back;

But it ends up looking like a "Draw this in your style," challenge and kinda proves ai's point. The whole 'stealing' ai is inherently using it, which I agree, other than the colouring or even painting style, the composition of the girl/cat I wouldn't copy that as it proves that I need the robot to think for me, but others think this is a way to 'stick it to the ai'

Personally, I would agree with Josh if it was using the generated image just for colouring, pallete ideas but using it for hair when he couldn've done that himself AND using Midjourney, putting money into a program that actually takes images from artists and using that stolen work for profit? "It's not stealing-"

It is, take a look at the 3-d model in CSP.

wonder why no one loses their shit when people trace these models for their generic webtoons? Because someone was payed for their time to create/code that model, it was put in there with CSP consent. Ai art, for me is like a Well that is open to the public, but the water people are taking from this Well is full of worms and parasites (aka. other peoples art) How can an artist use this when they are aware other artists work is in there, may as well trace and plagarise - the program is infested with other peoples work and if there was an isolated program I could use to train my own artwork then sure, but knowing that loish art is in there, has been trained it'll make all of my 'work' nothing but frankenstein abominations. I can't say with my full chest thats 'my' work.

And like this parasitic water, because it's free to the public people think I'm against free water and trying to force people to buy bottled and people are even going the Well water is more healthy for you.

All I've seen of ai, is done nothing but bog down platforms, it's a slog to find people who are real, not bots, not content farmers, not get rich quick pyramid scheme idiots, but people, communities.

Art is to 'connect' with people, if you couldn't be bothered to draw it, why should I be bothered to look at it, it tells me nothing about a person other than they have no passion.

[ciswomenofficial]

No matter whether AI does a good job at what it does or not, rage against the technology isn’t going to solve the labor issues. It isn’t going to stop artists and writers from losing their jobs. Whether it’s able to do the job adequately or not, it will be used to replace that job if all we do is protest against the technology itself. It will similarly do so if we rally around liberal principals like “copyright” and “intellectual property.” Most of the “intellectual property” belongs to conglomerates, not to intellectual laborers. The same if you misrepresent AIs water and energy consumption. None of these will get you anywhere productive, and some of them (“intellectual property”) even show a petty-bourgeoisie outlook.

Of course we should be concerned about small independent artists having their lives be less stable—even if it’s not our chief concern. Whether someone is an artisan proper or has some semi-proletarian characteristics (that is if they are working a side hustle to get by while they work on their passion project) that doesn’t mean we shouldn’t take their interests into account, but it does need to come from an industrial proletarian perspective first and foremost. Who will get laid off if AI technology is adopted? Who will lose their income? These are far more primary than independent artists.

The industrial proletariat is the most concentrated and unified force we have. It is this force with which we can fight and win. If we want to fight and win on the subject of AI, it must be fought with the organized industrial proletariat. Unions represent one lower form of this fight. The higher form of this fight is a political leadership that can lead to political reforms and regulations being adopted by the government as concessions, or perhaps even AI and other means of production being seized by the proletariat and regulated and technologically improved by the proletariat.