This. It's why AI animation fricking sucks, too, because on every frame details appear and disappear like crazy and it's just seizurebait.
If they manage to work that kink out like they did for the fricked up faces and fingers AI did last year, there might be something to it.
Not counting on it. Since big tech bought out all those AI research firms I expect AI to have the same fate VR did after the Zucc bought everyone out and snuffed them. They'll keep the cool datahoarding tech for themselves and we'll get frickall.
SIGH, if I wanted to see low quality uncanny images of Power Girl I'd generate them myself with the exact same tiny effort it took you to, and be able to adjust the prompt to my own tastes.
Enjoy your proompting WITHOUT annoying everyone else with its lackluster output.
It also can't really do interesting angles or extrapolate meaning from prompting tags it doesn't have coded into it. I spent the past day trying to get it to generate the loose fabric of an untied bikini but after six different models and hundreds of attempts I'm just giving up and doing it by hand.
For now AI art has this year alone improved massively and it will keep getting better. And the rate AI can pump out art demolishes human artists. Even if only 1 out of 100 is good, when the AI can pump out a thousand in the time it takes human to do one, the AI wins.
Look at the stuff AI was turning out at the start of the year, it was pretty bad. In less then a year it gotten to the point it can reliably turn out pretty decent art. Give it another year it and will be good art. Give it 5 years and high quality art will be the AI norm.
9 months ago
Anonymous
NTA but I am comparing it to the start of the year and even to last year and the truth is that while it's improved in some ways in terms of fidelity and detail, it has remained stagnant in the important ways like layout and dynamism. Fancier pinups are still just pin ups and yeah that will threaten a very specific subset of artists (maybe) but you can't tell a story with just dead eyed portraits of waifus. Eventually they need to do stuff like hold objects or eat or interact with each other and AI still struggles with that shit on a regular basis.
For the innovation to be an actual threat it'd need to be guided by those same artists it seeks to subsume because the people satisfied with increasingly better looking pin ups are demonstrably staying in that lane. I haven't seen anything coming out of the new SDXL model but "same thing as last month but with more pixels". Techbros surprisingly don't understand that new CONCEPTS is what makes art worthwhile not just better execution. >But in ten years!
We can revisit this conversation then. Until then it's futurist trolling is still just trolling.
You're basically saying that AI is the McDonald's of the artworld
It is mcdonalds in a sense that it's easy and unoriginal and can be mass-produced. Still leaves demand for "good food"
Some ai images can look less like generic aislop than op pic for sure though
9 months ago
Anonymous
In the sense that it's mass produced inorganic shit that while not bad per se, you'll never think about it again when it's been consumed.
Keeping up with this moronic food analogy; you can go to a steakhouse that sells escargot and vintage wine (high quality art) or you can just go to a hotdog cart manned by one person (amateur art made by beginners)
That's a pretty good take. It will absolutely replace the artists that are currently "the McDonald's of the art world" while artists with actual talent will still be able to find work.
It's a natural human ability. Same way gaydar and transclocking works.
I can't fap to this
It's OK. I'll fap twice for you.
>Why regret your work being seen by people? Anyone who knows jack shit about anything in the sphere that your shit took serious effort and wasn't just some prompt flinger shitting all over the place.
Which is next to nobody here. I don't like inadvertently aiding shitposters.
>Since the people making the claims know nothing about the tools, much less the theory and the methodology, and are more interested in ignorant speculative shitposting, it's best not to take them too seriously.
Except in this case they're kind of right about the limitations of the technology. It doesn't actually "know" what hands are in the same way you or I do. It's taught that there's a collection of pixels at the end of another collection of pixels but the idea that "this doesn't bend that way" isn't a thing it understands because it doesn't actually know what those pixels *are*. That much is inherently true.
>That much is inherently true.
For many current engines, but it's not inherent. Look at the shit that takes poser, and mocap input. Those definitely have a concept of anatomy.
>For many current engines, but it's not inherent.
When I said "inherently true" I meant the statement that AI isn't truely intelligent is true. And that is still the case. >Look at the shit that takes poser, and mocap input. Those definitely have a concept of anatomy.
Yes, controlnets like that have data being fed into them that they mirror back at you. That's not the same as having an intuitive concept of anatomy like that says >user is asking for a hand. hands have five fingers.
It can certainly trace a picture you give it of a hand via control net but it's not actually *thinking* about it in te same way you or I would do so and that's an important distinction to make so as to lower user expectations as to what this shit can actually do.
That's what happens when they all have a standardized "interpretation" of what constitutes beauty. It's like how a year ago every girl was Emma Watson. The only way to change this is diversity of training samples but diversity is a no no word for the kind of person that's in this to make artists seethe so we're just gonna have to deal with the sameface forever.
>Might have helped that the post has the words AI in it
That's totally unnecessary. Once you've seen a few hundred AI works you start to recognize them from the thumbnail. AIshit has a certain je ne quois that's instantly recognizable.
Alegria and CalArts toons already might as well be AI >Lifeless, souless, samey looking crap made by machines, copying other things that copy other things that copy other things that copy other things that copy other things that copy other things, without any human input or concern for beauty or intent.
The only difference is that this is made by a computer, whereas the machines that make Calarts and corpo-slop garbage are flesh and blood, but they still think like machines.
The npc meme has never been more accurate than it is now.
Why does Cinemaphile even get AI threads? Especially since they hardly ever discuss AI in comics or cartoons, but simply in horny illustration slop? Wouldn't this fit elsewhere better, like an art board or something? Or do those have jannies and communities that actually give a frick and won't bite your shitty bait?
/Ic/ still gets a load of AI threads despite them being deleted and made as bait threads
Cinemaphile really should be the place for them but no one really cares and people would rather show off their super special original generated character than actually discuss the ethics or future of this stuff
It's all so tiresome
>AIgays preach that AI is simply a tool and a technological advancement >Won't ever post on the board FOR technology
Uh huh.... We all know they really see it as easy, mass produced art slop for their dopamine fried brains that constantly need new stuff to look at no matter how shit
It also can't really do interesting angles or extrapolate meaning from prompting tags it doesn't have coded into it. I spent the past day trying to get it to generate the loose fabric of an untied bikini but after six different models and hundreds of attempts I'm just giving up and doing it by hand.
this. it's too frustrating. manual drawing has more control than generating ai images
Because threads that get replies get made again. Anons flock to these threads to complain and the person that made the thread is encouraged to make more.
As an artist who can now make any kind of reference image he wants I laugh at all you losers who thought you were irreplaceable.
Art is about expression and creativity, it's not just some job to stroke your ego.
Art itself is the reward, and this is why AI will replace you.
You. Forgot.
How is the AI supposed to replace artists if it can't event generate images without stealing existing art? Hell, you can't even make porn of your favorite cartoon character if an artist doesn't invent your favorite cartoon character in the first place. You people are so moronic. I know the NFT scam fell, but you don't immediately need to switch gears to another grift. You COULD try making something yourself and maybe be a net positive force on the world instead of what you are now.
I'm not disagreeing. At the end of the day stable diffusion is just a gacha machine for making waifu pics. Loras increase your odds but you've still got to roll the dice for anything halfway decent.
I'm sure someone already thought about this, but if Megas and an Autobot get together, would Megas be considered a sex toy? I know Coop would be the one controlling Megas, but from his perspective, he's just playing a game.
My stupid ramble gave me an idea involving Kiva, a modified Megas wienerpit, and Arcee/Windblade. The question is, can I find an HD version of Megas XLR?
The general plan is outlined. Now I need a Miko and Airachnid lora. The hardest part will be the oiled pistons, and I don't think reaction shots are necessary.
It'll be a fun little project, and since these bait threads are a weekly occurrence, I can maintain a weekly journal on Cinemaphile without bothering real threads.
Alright anon I really wanna figure out how to do this stuff, All I've ever used is Novel AI. What is the best place to get started on Stable Diffusion?
That's what happens when they all have a standardized "interpretation" of what constitutes beauty. It's like how a year ago every girl was Emma Watson. The only way to change this is diversity of training samples but diversity is a no no word for the kind of person that's in this to make artists seethe so we're just gonna have to deal with the sameface forever.
Most of the pressure I've seen is a complete fear of the engines being used to make shit that seems "racist".
>Most of the pressure I've seen is a complete fear of the engines being used to make shit that seems "racist".
The gulf between ethnic features being seen as genuine or racist can only be viewed with an electron microscope
I'm not surprised. If Blizzard's artists had to make all their Asian women white to avoid accusations of racism, what chance does an AI have?
You know what, that's fair.
I suppose if you're too lazy to check the catalog then an AI thread is actually a good place for you.
9 months ago
Anonymous
>Branch of conversation was started due to a question as to why sameface with AI >Suggestion that it's because the creators are scared of making racist caricatures >"Uhhh why didn't you search the entire catalog for a thread on page 9 about real artists being accused of giving white features to black people?"
Seriously, go frick yourself
9 months ago
Anonymous
I thought AI gays wanted to be considered real artists though?
>The gulf between ethnic features being seen as genuine or racist can only be viewed with an electron microscope
It's worse. It's guaranteed to change.
the problem with ai, and this will always to some degree be a problem, is that the machine doesn't understand the concepts of what it's creating, it's just replicating patterns
that's why the hands are always jacked up, it's cause the machine knows that more often than not human figures have an appendage with wiggly fingers at the end but what those fingers are doing are always contextual. so it's not making a decision to specifically draw those fingers doing something, it's just regurgitating the most common positions that it has seen. it also doesn't understand perspective or for shortening, so it's never truly sure how many fingers are needed to be seen at any given time. hell it doesn't know what fingers are, just that they are supposed to be there in some sort of arrangement
the reason boobs and anime faces turn out ok sometimes is the millions of pics of boobs and anime faces that look exactly the same. but again that's only sometimes, cause the machine doesn't know what boobs are, or that the eyelashes go above the eye, its just spitting out compiled data
It's hilarious seeing them all cling to what is essentially a dead end. The only way generative "AI" is ever going to truly reach a state where it becomes indistinguishable from a real person's work is if it becomes truly intelligent which we are still decades away from if even that. Right now we're just brute forcing the data.
You know jack shit about AI. What is happening is that AI based tools are taking increasing amounts of the workflow, and accelerate things for existing artists. Less photoshop and lightbox abuse by porno specialist cover artists who's names I don't need to mention, replaced with generative AI doing parts of the workload. Shit like colorizing. Backgrounds. Ect.
You don't need a generalized artificial inteligence to get all the way down so that some fiver doofus can prompt things into existence. In fact, I'm betting you can't identify the AI generated portions of existing works with any consistency.
>muh workflow
It still looks like shit and even the barest amount of scrutiny lays it out bare for what it is. Generative AI is a dead end and has no place in the art world outside of grifters and charlatans.
How about you show me one so I can tell you it looks like crap?
9 months ago
Anonymous
You made the moronic fricking claim. Back it up or frick off. Cinemaphile enjoys shitting on Greg Land, and has plenty of bad art threads, so I'm sure it will do just fine, even it it's own thread..
9 months ago
Anonymous
You're free to post any work you desire and I will judge it, I have no interest in trawling for AIslop
>. Shit like colorizing. Backgrounds.
All the backgrounds posted ITT have significant issues though. And speaking as someone that's tried using this tech for colorizing it's kind of shit at that too. Controlnets don't respect line art nearly as much as they should even if I set it to pixel perfect and forget about doing anything particularly high res. Then you're just asking for a CUDA error.
Absolutely. But you can see how generating a background, or generating 30 and picking one as a starting point is much more manageable.
There's no rule that says you have to do it all in one prompt. Getting some of the other elements via generative fill and iteration is still a big timesaver.
The lineart issue is very real, and will is being worked on by some sort of line extraction method like XDoG, SketchKeras or simplification. A better interface for hinting is being worked on. That's also not counting people going with line art methods that are AI friendly to work with.
Development on both these areas doesn't require generalized AI.
I think when people say "generalized AI" they're really talking about intuition in such a way that the base technology isn't really set up for. Attaching meaning to tags to tags is still being done by the trainer, not the machine itself.
>You know jack shit about AI. What is happening is that AI based tools are taking increasing amounts of the workflow, and accelerate things for existing artists. Less photoshop and lightbox abuse by porno specialist cover artists who's names I don't need to mention, replaced with generative AI doing parts of the workload. Shit like colorizing. Backgrounds. Ect. > >You don't need a generalized artificial inteligence to get all the way down so that some fiver doofus can prompt things into existence. In fact, I'm betting you can't identify the AI generated portions of existing works with any consistency.
Any amount of AI is fine if the person who's publishing it took the time to edit the kinks. Otherwise it's just art the computer did, no credit goes to the person whose only contribution was typing in keywords that any monkey could.
AI can be neat for stuff that's kind of niche or if a thread doesn't have a bunch of drawgays. I'm not into companies using it, but for porn or more minor shit like this, it's fine.
GG gets decent art, so maybe not the best example tho
>but for porn or more minor shit like this, it's fine
Frick that, it's already starting to flood places I frequent making it progressively more annoying to find shit I enjoy.
Speaking as the anon that made that I kind of regret uploading it because now so many people post it without the context I provided as to how huge a pain it in the ass it was to get those results.
Why regret your work being seen by people? Anyone who knows jack shit about anything in the sphere that your shit took serious effort and wasn't just some prompt flinger shitting all over the place.
I think when people say "generalized AI" they're really talking about intuition in such a way that the base technology isn't really set up for. Attaching meaning to tags to tags is still being done by the trainer, not the machine itself.
Since the people making the claims know nothing about the tools, much less the theory and the methodology, and are more interested in ignorant speculative shitposting, it's best not to take them too seriously.
Credit to the AI... at least it knows where to cut the leotard... mostly.
They learned from the best.
I'm sure someone already thought about this, but if Megas and an Autobot get together, would Megas be considered a sex toy? I know Coop would be the one controlling Megas, but from his perspective, he's just playing a game.
That is a GREAT fricking question, and if I bothered with any social media, I would ask fred perry.
>Why regret your work being seen by people? Anyone who knows jack shit about anything in the sphere that your shit took serious effort and wasn't just some prompt flinger shitting all over the place.
Which is next to nobody here. I don't like inadvertently aiding shitposters.
>Since the people making the claims know nothing about the tools, much less the theory and the methodology, and are more interested in ignorant speculative shitposting, it's best not to take them too seriously.
Except in this case they're kind of right about the limitations of the technology. It doesn't actually "know" what hands are in the same way you or I do. It's taught that there's a collection of pixels at the end of another collection of pixels but the idea that "this doesn't bend that way" isn't a thing it understands because it doesn't actually know what those pixels *are*. That much is inherently true.
You cannot control your work very much once it leaves your hands. Which is an irony of AI production methods. You made fricking awesome shit, and don't let any butthole who wants to shit on all AI art, or some simp scammer who wants to sell shit.
I hope you make more shit. Not necessarily AI or golden girl. But if you don't I want you to know that people appreciate what you made.
That's just it. I'd also rather shit on AI art. Or at least, I'd rather people have some perspective as to how gimmicky it is.
I can actually draw AND use stable diffusion and the latter's limitations are a huge issue still. But when people post the end result of generations it somehow gives the impression that it's the result of a first attempt, when the behind every single one of those pictures was 50-60 failed eldritch looking shit generations that ended up in the recycle bin.
I'm going to suggest a piece anon. Remember the Atom Eve fight from her prequel? Imagine a Golden Girl seeing pictures hundereds of misformed clones made by some mad scientist.
>What are these? >They're you. Other dimensions other attempts.
All failures. They had to be born and die so you could live. >WHAT DID YOU DO? >I DID WHAT HAD TO BE DONE. I DID THE NEEDUFL.
Kind of rushed the post button, but the core idea was to use all the messed up iterations so that people don't see what you did as just "press button, get art" .
I fail to see how what I said warrant that kind of response. All I said was your image was neat to see since the thread didn't have a lot of people actively drawing in it. How that makes me as
This is why you don't post good intentions on Cinemaphile, because this place has evil people come to it.
said, an "evil" person, I don't know. But I'll refrain from giving you positive attention in the future if you're this against it
AI's kinda in the same block as any other digital tool. Just like digital animation and tweening or stuff like rotoscoping or interpolation, it'll cheapen the look of things across the board, but people will generally have persistent hate for the laziest of animation, or that which otherwise looks like they just used what was out of the box. People who actually use it with their own skills could do quite well.
I feel like that perception would be more popular if the people who touted AI as a tool didn't also immediately leap to "this is replacing artists, you don't need any old skill just being a prompt jockey, look here's this masterpiece that I was able to make with just a simple prompt (shows forgettable, overprocessed cowboy shot that looks like it's from somebody's forgettable Artstation profile).
I also think most artistic endeavors would still benefit from the manual touch and still having a staff of people. You listen or read to any behind the scenes stuff and it's clear that the team as a whole can have important artistic input. Voice acting being replaced by AI is the big one nowadays, but anyone who's actually heard about the process and efforts of the actors knows that how they think, improvise, or even just exist as a person independent from the director is often invaluable. Projects that think that they can ditch those things but ultimately suffer in ways they don't notice is when it's too much AI
You're completely right, but I doubt you'll get any decent discussions on here. People just either wanna be all for AI or all against it. There's nothing new being said in this thread
I get you. I must admit, I sorta enjoy AI now, but I've been too stressed out about how it won't get the exact results that I want. It gets especially difficult if you want to involve more than one subject. I might go back into AI generated art later but now I'm just doing manual. The social stigma surrounding AI art doesn't help, either
They should considering he draws copyrighted characters and has copycat hentai artists copying him and now a a.i. lora to imitate that style but can be mixed in with other loras
Loras themselves are plagiarism. They are also illegal since you remove artists' watermarks to create your loras. Blatant theft.
9 months ago
Anonymous
some keep the watermark and insert patreon watermarks and you have to edit them out and put them in the negatives
9 months ago
Anonymous
Any messing with watermarks is an infringing behavior in violation of DMCA Section 1202 and makes you liable to be pursued legally.
9 months ago
Anonymous
Gasp
9 months ago
Anonymous
why can't you do anything without literally stealing other people's art?
9 months ago
Anonymous
It's not stealing if nobody can claim it
9 months ago
Anonymous
the artist can indeed claim the training data
and the EU is forcing ai generators to release all copyrighted training data
9 months ago
Anonymous
which artist can claim this pic and would go though the litigation?
9 months ago
Anonymous
All artists are already going through the litigation of getting their art removed from the training data.
https://haveibeentrained.com/
9 months ago
Anonymous
but you can just save pics and train your own lora
so how does slow litigation to stop a website that will be cloned stop anyone from posting a a.i. gen pic of Tifa vs a drawn one on Cinemaphile?
9 months ago
Anonymous
>training on copyrighted images
That's the no-no. Any website that does that is illegal. You creating your own lora on copyrighted images makes you liable to be pursued legally.
9 months ago
Anonymous
So how does one go after a person who trained a powergirl lora using hentai and rule 34 artists who also don't have copyright claim to the material?
I made a powergirl lora in Shademan style, do you go after me or keep fricking his corpse?
and if you go after me do you go after all proper rule 34 artists and hentai artist who draw powergirl and have patreons and stuff but have no legal reason to draw her?
9 months ago
Anonymous
Shadman can go after you since his style includes his original stuff
9 months ago
Anonymous
but all his material is already a copyright infringement like most artists
9 months ago
Anonymous
Not at all. He doesn't just draw fanart
9 months ago
Anonymous
he would have to go though a hell of a lot of legal work to doge the Dagne Keen fiasco and all the work he did with e-girls from drifferent properties with no claim to take legal action on a anon positng on Cinemaphile, I suggest you tweet him it's going on in real time.
9 months ago
Anonymous
dude there's people who literally state >LoRA trained on Lois van Baarles works (Loish)
that just want to literally replace the artist
9 months ago
Anonymous
that's why those models/lora are very iffy to me. if it's just the "mix" variety, i'm alright with them, since their sources are very diverse
9 months ago
Anonymous
those aren't "iffy", they are literal illegal shit explicitly trained on illegal material since Loish never said you can use her original art
>the ai can only do ATSV Spider-Gwen porn bate lips now
I wish this tech was around when the internet went insane over Lady Dimitrescue, all the generations would have looked taller and mommy-ish.
Even without looking at the fricked up clocktower I instantly knew it was AI.
Might have helped that the post has the words AI in it, aspergers-kun
Too much A, too little I.
not funny
I don't remember asking for your pronouns.
The one thing ai can't do properly yet is consistent designs, look at the cape
This. It's why AI animation fricking sucks, too, because on every frame details appear and disappear like crazy and it's just seizurebait.
If they manage to work that kink out like they did for the fricked up faces and fingers AI did last year, there might be something to it.
Not counting on it. Since big tech bought out all those AI research firms I expect AI to have the same fate VR did after the Zucc bought everyone out and snuffed them. They'll keep the cool datahoarding tech for themselves and we'll get frickall.
Yes, it’s so fricking over.
I like how it looks like it incorporated the belly button in the zipper
SIGH, if I wanted to see low quality uncanny images of Power Girl I'd generate them myself with the exact same tiny effort it took you to, and be able to adjust the prompt to my own tastes.
Enjoy your proompting WITHOUT annoying everyone else with its lackluster output.
It also can't really do interesting angles or extrapolate meaning from prompting tags it doesn't have coded into it. I spent the past day trying to get it to generate the loose fabric of an untied bikini but after six different models and hundreds of attempts I'm just giving up and doing it by hand.
Without the data to scalp, it will never be able to remix pixels into new consistent imagery.
ai art has what i personally call 'cosplay syndrome'. most of the time they never look like the character theyre supposed to be
For now AI art has this year alone improved massively and it will keep getting better. And the rate AI can pump out art demolishes human artists. Even if only 1 out of 100 is good, when the AI can pump out a thousand in the time it takes human to do one, the AI wins.
You're basically saying that AI is the McDonald's of the artworld
Look at the stuff AI was turning out at the start of the year, it was pretty bad. In less then a year it gotten to the point it can reliably turn out pretty decent art. Give it another year it and will be good art. Give it 5 years and high quality art will be the AI norm.
NTA but I am comparing it to the start of the year and even to last year and the truth is that while it's improved in some ways in terms of fidelity and detail, it has remained stagnant in the important ways like layout and dynamism. Fancier pinups are still just pin ups and yeah that will threaten a very specific subset of artists (maybe) but you can't tell a story with just dead eyed portraits of waifus. Eventually they need to do stuff like hold objects or eat or interact with each other and AI still struggles with that shit on a regular basis.
For the innovation to be an actual threat it'd need to be guided by those same artists it seeks to subsume because the people satisfied with increasingly better looking pin ups are demonstrably staying in that lane. I haven't seen anything coming out of the new SDXL model but "same thing as last month but with more pixels". Techbros surprisingly don't understand that new CONCEPTS is what makes art worthwhile not just better execution.
>But in ten years!
We can revisit this conversation then. Until then it's futurist trolling is still just trolling.
Well...He's not wrong.
It is mcdonalds in a sense that it's easy and unoriginal and can be mass-produced. Still leaves demand for "good food"
Some ai images can look less like generic aislop than op pic for sure though
In the sense that it's mass produced inorganic shit that while not bad per se, you'll never think about it again when it's been consumed.
Keeping up with this moronic food analogy; you can go to a steakhouse that sells escargot and vintage wine (high quality art) or you can just go to a hotdog cart manned by one person (amateur art made by beginners)
That's a pretty good take. It will absolutely replace the artists that are currently "the McDonald's of the art world" while artists with actual talent will still be able to find work.
Her costume is messed up around the belt.
>Might have helped that the post has the words AI in it, aspergers-kun
Dude, my Sharingan sees AI posts in an instant. I don't understand why I've developed this capability or what purposes it serves.
It's just basic pattern recognition. Turns out when you flood spaces with a bunch of identical looking shit everyone is going to twig on to it.
It's the lighting for me. Noticeable even from a thumbnail.
For me, it was her thighs. Even in the thumbnail, her things look whack as frick.
It's a natural human ability. Same way gaydar and transclocking works.
It's OK. I'll fap twice for you.
>That much is inherently true.
For many current engines, but it's not inherent. Look at the shit that takes poser, and mocap input. Those definitely have a concept of anatomy.
>For many current engines, but it's not inherent.
When I said "inherently true" I meant the statement that AI isn't truely intelligent is true. And that is still the case.
>Look at the shit that takes poser, and mocap input. Those definitely have a concept of anatomy.
Yes, controlnets like that have data being fed into them that they mirror back at you. That's not the same as having an intuitive concept of anatomy like that says
>user is asking for a hand. hands have five fingers.
It can certainly trace a picture you give it of a hand via control net but it's not actually *thinking* about it in te same way you or I would do so and that's an important distinction to make so as to lower user expectations as to what this shit can actually do.
For me its super noticeable in the faces.
All AI Images seem to draw the same style of face
That's what happens when they all have a standardized "interpretation" of what constitutes beauty. It's like how a year ago every girl was Emma Watson. The only way to change this is diversity of training samples but diversity is a no no word for the kind of person that's in this to make artists seethe so we're just gonna have to deal with the sameface forever.
>Might have helped that the post has the words AI in it
That's totally unnecessary. Once you've seen a few hundred AI works you start to recognize them from the thumbnail. AIshit has a certain je ne quois that's instantly recognizable.
>Once you've seen a few hundred AI works you start to recognize them from the thumbnail.
Oy vey, very antisemitic!
Which AI is this?? Model and Prompts???
Credit to the AI... at least it knows where to cut the leotard... mostly.
You don't watch much porn... do you?
I took that picture because it's what's best in life
Dunno what your question really means
I thought you may have been disagreeing with the anon. If not, my bad fellow tight-cut leotard connoisseur
AI will be used to make endless Alegria and CalArts toons.
Alegria and CalArts toons already might as well be AI
>Lifeless, souless, samey looking crap made by machines, copying other things that copy other things that copy other things that copy other things that copy other things that copy other things, without any human input or concern for beauty or intent.
The only difference is that this is made by a computer, whereas the machines that make Calarts and corpo-slop garbage are flesh and blood, but they still think like machines.
The npc meme has never been more accurate than it is now.
Because corpos buy that, so they will buy the same shit with AI. Ten times the output, Zero times the price.
pic related is too AI
Why does Cinemaphile even get AI threads? Especially since they hardly ever discuss AI in comics or cartoons, but simply in horny illustration slop? Wouldn't this fit elsewhere better, like an art board or something? Or do those have jannies and communities that actually give a frick and won't bite your shitty bait?
Those boards are lower traffic, so Cinemaphile is probably the largest board who jannies seem to allow them to get (you)s
/Ic/ still gets a load of AI threads despite them being deleted and made as bait threads
Cinemaphile really should be the place for them but no one really cares and people would rather show off their super special original generated character than actually discuss the ethics or future of this stuff
It's all so tiresome
>AIgays preach that AI is simply a tool and a technological advancement
>Won't ever post on the board FOR technology
Uh huh.... We all know they really see it as easy, mass produced art slop for their dopamine fried brains that constantly need new stuff to look at no matter how shit
/ai/ should get its own board
this. it's too frustrating. manual drawing has more control than generating ai images
Because threads that get replies get made again. Anons flock to these threads to complain and the person that made the thread is encouraged to make more.
>Cinemaphile will pretend this art is bad because it's le AI
Who is the original artist the AI based the shading on
When low-wage physical labor still requires humans while robots make art
Robots cost money.
Any AI is too much
This AI sucks ass no matter what. I've never seen a good piece of AI 'art' ever.
it would be better if every ai picture you guys post wasn't in the patreonslop art style
>it can make ANYTHING!
>proceeds to make the same shit everyone has already seen for 5-10 years
what is up with the BG?
As an artist who can now make any kind of reference image he wants I laugh at all you losers who thought you were irreplaceable.
Art is about expression and creativity, it's not just some job to stroke your ego.
Art itself is the reward, and this is why AI will replace you.
You. Forgot.
I need to eat.
>As an artist who can now make any kind of reference image he wants I laugh at all you losers who thought you were irreplaceable.
Post work.
>Art is about expression and creativity
So how are you expressing or being creative by outsourcing those to elements to a machine?
How is the AI supposed to replace artists if it can't event generate images without stealing existing art? Hell, you can't even make porn of your favorite cartoon character if an artist doesn't invent your favorite cartoon character in the first place. You people are so moronic. I know the NFT scam fell, but you don't immediately need to switch gears to another grift. You COULD try making something yourself and maybe be a net positive force on the world instead of what you are now.
Personally, I think AI is fun.
Damn the Tetris world is getting fricked up behind that b***h.
very ugly shoes and paper thin knees
0/10
I’m sorry you feel that way.
I can't fap to this
So there is a setting to dress anyone like powergirl?
There are probably a couple loras for her. For character specific stuff using loras instead of models seems to be the way to go.
It still looks like shit, it just mildly resembles the character more. All this
>mu lora
stuff is just a cope.
So when you look at this, it only mildly reminds you of Power Girl?
Some generic superhero trash for the most part. Unlike you, I'm not so easily entertained by whatever slop is put in front of me.
But Anon, Power Girl is DCSlop personified.
So you're saying Slop2 is an improvement?
I'm not disagreeing. At the end of the day stable diffusion is just a gacha machine for making waifu pics. Loras increase your odds but you've still got to roll the dice for anything halfway decent.
Right on the money. Have a different Power Gwen
>Gwen Tennyson if she decided to be a legacy character, rather than her own hero
You know... I can get behind the idea.
>How AI is too AI?
too AI? What?
Did something happen again? Why is pajeet sperging out this time?
>How AI is too AI?
When it looks like shit
>When it looks like shit
pic related?
why are you gae?
why do you have such low stand
ards?
So always?
Okay, now post an example
At least try and make it look western.
These both look like doodoo ass
Giant woman!
Honestly I love to see some AI art of adult pornographic anime.
Since this is a shitpost thread, I'm going to have my fun. And before you reply to me, look at the page count.
All I wanna be is someone who gets to see a giant woman.
I'm sure someone already thought about this, but if Megas and an Autobot get together, would Megas be considered a sex toy? I know Coop would be the one controlling Megas, but from his perspective, he's just playing a game.
My stupid ramble gave me an idea involving Kiva, a modified Megas wienerpit, and Arcee/Windblade. The question is, can I find an HD version of Megas XLR?
Damn, the best I could find was 480p. Probably for the best. Some ideas shouldn't leave one's head.
On the other hand, I can use Miko, two wireless pistons, Arcee, and the arachnid lady...
The general plan is outlined. Now I need a Miko and Airachnid lora. The hardest part will be the oiled pistons, and I don't think reaction shots are necessary.
It'll be a fun little project, and since these bait threads are a weekly occurrence, I can maintain a weekly journal on Cinemaphile without bothering real threads.
So Step 1: Build a Miko dataset.
I can't wait until they start animating with ai.
Why are all AI females turning into Gwen?
This here is with a Gwen lora added.
Are those wooden blocks or buildings? what the hell?
I don't care for this art style and Power Girl's costume is stupid.
I want a Rio Morales spider-gwen
pointy cape makes no sense
Alright anon I really wanna figure out how to do this stuff, All I've ever used is Novel AI. What is the best place to get started on Stable Diffusion?
Frick off to the poojeet Cinemaphileenerals if you want to learn how to do the needful
I shit ton of filtering.
By visiting Cinemaphile.
Most of the pressure I've seen is a complete fear of the engines being used to make shit that seems "racist".
>Most of the pressure I've seen is a complete fear of the engines being used to make shit that seems "racist".
The gulf between ethnic features being seen as genuine or racist can only be viewed with an electron microscope
I'm not surprised. If Blizzard's artists had to make all their Asian women white to avoid accusations of racism, what chance does an AI have?
There's already a thread for this bullshit use that one instead.
Not everyone monitors the entire site constantly, schizo
Besides that wasn't an AI thread
You know what, that's fair.
I suppose if you're too lazy to check the catalog then an AI thread is actually a good place for you.
>Branch of conversation was started due to a question as to why sameface with AI
>Suggestion that it's because the creators are scared of making racist caricatures
>"Uhhh why didn't you search the entire catalog for a thread on page 9 about real artists being accused of giving white features to black people?"
Seriously, go frick yourself
I thought AI gays wanted to be considered real artists though?
>The gulf between ethnic features being seen as genuine or racist can only be viewed with an electron microscope
It's worse. It's guaranteed to change.
Unironic question: how do you manage to create those images with AI? Every time I try it I only create abominations.
When the audience is only AI.
the problem with ai, and this will always to some degree be a problem, is that the machine doesn't understand the concepts of what it's creating, it's just replicating patterns
that's why the hands are always jacked up, it's cause the machine knows that more often than not human figures have an appendage with wiggly fingers at the end but what those fingers are doing are always contextual. so it's not making a decision to specifically draw those fingers doing something, it's just regurgitating the most common positions that it has seen. it also doesn't understand perspective or for shortening, so it's never truly sure how many fingers are needed to be seen at any given time. hell it doesn't know what fingers are, just that they are supposed to be there in some sort of arrangement
the reason boobs and anime faces turn out ok sometimes is the millions of pics of boobs and anime faces that look exactly the same. but again that's only sometimes, cause the machine doesn't know what boobs are, or that the eyelashes go above the eye, its just spitting out compiled data
It's hilarious seeing them all cling to what is essentially a dead end. The only way generative "AI" is ever going to truly reach a state where it becomes indistinguishable from a real person's work is if it becomes truly intelligent which we are still decades away from if even that. Right now we're just brute forcing the data.
You know jack shit about AI. What is happening is that AI based tools are taking increasing amounts of the workflow, and accelerate things for existing artists. Less photoshop and lightbox abuse by porno specialist cover artists who's names I don't need to mention, replaced with generative AI doing parts of the workload. Shit like colorizing. Backgrounds. Ect.
You don't need a generalized artificial inteligence to get all the way down so that some fiver doofus can prompt things into existence. In fact, I'm betting you can't identify the AI generated portions of existing works with any consistency.
>muh workflow
It still looks like shit and even the barest amount of scrutiny lays it out bare for what it is. Generative AI is a dead end and has no place in the art world outside of grifters and charlatans.
I think you're hot air. Show me an example of generative fill being obviously detectable and looking like shit in a pro project.
How about you show me one so I can tell you it looks like crap?
You made the moronic fricking claim. Back it up or frick off. Cinemaphile enjoys shitting on Greg Land, and has plenty of bad art threads, so I'm sure it will do just fine, even it it's own thread..
You're free to post any work you desire and I will judge it, I have no interest in trawling for AIslop
post work
>. Shit like colorizing. Backgrounds.
All the backgrounds posted ITT have significant issues though. And speaking as someone that's tried using this tech for colorizing it's kind of shit at that too. Controlnets don't respect line art nearly as much as they should even if I set it to pixel perfect and forget about doing anything particularly high res. Then you're just asking for a CUDA error.
Absolutely. But you can see how generating a background, or generating 30 and picking one as a starting point is much more manageable.
There's no rule that says you have to do it all in one prompt. Getting some of the other elements via generative fill and iteration is still a big timesaver.
The lineart issue is very real, and will is being worked on by some sort of line extraction method like XDoG, SketchKeras or simplification. A better interface for hinting is being worked on. That's also not counting people going with line art methods that are AI friendly to work with.
Development on both these areas doesn't require generalized AI.
I think when people say "generalized AI" they're really talking about intuition in such a way that the base technology isn't really set up for. Attaching meaning to tags to tags is still being done by the trainer, not the machine itself.
>You know jack shit about AI. What is happening is that AI based tools are taking increasing amounts of the workflow, and accelerate things for existing artists. Less photoshop and lightbox abuse by porno specialist cover artists who's names I don't need to mention, replaced with generative AI doing parts of the workload. Shit like colorizing. Backgrounds. Ect.
>
>You don't need a generalized artificial inteligence to get all the way down so that some fiver doofus can prompt things into existence. In fact, I'm betting you can't identify the AI generated portions of existing works with any consistency.
in this kind of quoting you are not supposed to give the anon a (You). Better luck next time
Any amount of AI is fine if the person who's publishing it took the time to edit the kinks. Otherwise it's just art the computer did, no credit goes to the person whose only contribution was typing in keywords that any monkey could.
AI can be neat for stuff that's kind of niche or if a thread doesn't have a bunch of drawgays. I'm not into companies using it, but for porn or more minor shit like this, it's fine.
GG gets decent art, so maybe not the best example tho
>but for porn or more minor shit like this, it's fine
Frick that, it's already starting to flood places I frequent making it progressively more annoying to find shit I enjoy.
Speaking as the anon that made that I kind of regret uploading it because now so many people post it without the context I provided as to how huge a pain it in the ass it was to get those results.
I really like the one in her underwear so thanks for your snacrifice
Why regret your work being seen by people? Anyone who knows jack shit about anything in the sphere that your shit took serious effort and wasn't just some prompt flinger shitting all over the place.
Since the people making the claims know nothing about the tools, much less the theory and the methodology, and are more interested in ignorant speculative shitposting, it's best not to take them too seriously.
They learned from the best.
That is a GREAT fricking question, and if I bothered with any social media, I would ask fred perry.
>Why regret your work being seen by people? Anyone who knows jack shit about anything in the sphere that your shit took serious effort and wasn't just some prompt flinger shitting all over the place.
Which is next to nobody here. I don't like inadvertently aiding shitposters.
>Since the people making the claims know nothing about the tools, much less the theory and the methodology, and are more interested in ignorant speculative shitposting, it's best not to take them too seriously.
Except in this case they're kind of right about the limitations of the technology. It doesn't actually "know" what hands are in the same way you or I do. It's taught that there's a collection of pixels at the end of another collection of pixels but the idea that "this doesn't bend that way" isn't a thing it understands because it doesn't actually know what those pixels *are*. That much is inherently true.
You cannot control your work very much once it leaves your hands. Which is an irony of AI production methods. You made fricking awesome shit, and don't let any butthole who wants to shit on all AI art, or some simp scammer who wants to sell shit.
I hope you make more shit. Not necessarily AI or golden girl. But if you don't I want you to know that people appreciate what you made.
That's just it. I'd also rather shit on AI art. Or at least, I'd rather people have some perspective as to how gimmicky it is.
I can actually draw AND use stable diffusion and the latter's limitations are a huge issue still. But when people post the end result of generations it somehow gives the impression that it's the result of a first attempt, when the behind every single one of those pictures was 50-60 failed eldritch looking shit generations that ended up in the recycle bin.
>That's just it. I'd also rather shit on AI art
b-b-based!
I'm going to suggest a piece anon. Remember the Atom Eve fight from her prequel? Imagine a Golden Girl seeing pictures hundereds of misformed clones made by some mad scientist.
>What are these?
>They're you. Other dimensions other attempts.
All failures. They had to be born and die so you could live.
>WHAT DID YOU DO?
>I DID WHAT HAD TO BE DONE. I DID THE NEEDUFL.
Kind of rushed the post button, but the core idea was to use all the messed up iterations so that people don't see what you did as just "press button, get art" .
This is why you don't post good intentions on Cinemaphile, because this place has evil people come to it.
I fail to see how what I said warrant that kind of response. All I said was your image was neat to see since the thread didn't have a lot of people actively drawing in it. How that makes me as
said, an "evil" person, I don't know. But I'll refrain from giving you positive attention in the future if you're this against it
I like that she looks a little older
I'm not the sort to cry and b***h and claim illegality over that sort of thing, but it's closer to what I like
Can non-coom AI ever be useful for character art?
Depends on the use. This is still, in effect, a pin up.
All AI looks too AI by default
Moar images
i don't like how in the pictures you've posted one iris is at the side of an eye and the other is in the middle
makes her look moronic
They don't ever look at the details, that just want booba. Literally the "don't ask questions, just consume product" meme made manifest.
Any percentage above 0%
They couldn't at least fix the + to an X?
AI's kinda in the same block as any other digital tool. Just like digital animation and tweening or stuff like rotoscoping or interpolation, it'll cheapen the look of things across the board, but people will generally have persistent hate for the laziest of animation, or that which otherwise looks like they just used what was out of the box. People who actually use it with their own skills could do quite well.
I feel like that perception would be more popular if the people who touted AI as a tool didn't also immediately leap to "this is replacing artists, you don't need any old skill just being a prompt jockey, look here's this masterpiece that I was able to make with just a simple prompt (shows forgettable, overprocessed cowboy shot that looks like it's from somebody's forgettable Artstation profile).
I also think most artistic endeavors would still benefit from the manual touch and still having a staff of people. You listen or read to any behind the scenes stuff and it's clear that the team as a whole can have important artistic input. Voice acting being replaced by AI is the big one nowadays, but anyone who's actually heard about the process and efforts of the actors knows that how they think, improvise, or even just exist as a person independent from the director is often invaluable. Projects that think that they can ditch those things but ultimately suffer in ways they don't notice is when it's too much AI
You're completely right, but I doubt you'll get any decent discussions on here. People just either wanna be all for AI or all against it. There's nothing new being said in this thread
I'm all for it but I know I will never pay for it, well in the means of paying for pics, I will buy a pc.
I think it has it's use and I like the versatility
I get you. I must admit, I sorta enjoy AI now, but I've been too stressed out about how it won't get the exact results that I want. It gets especially difficult if you want to involve more than one subject. I might go back into AI generated art later but now I'm just doing manual. The social stigma surrounding AI art doesn't help, either
add "sketch" to your prompts
been messing around with hentai artist Loras, X-23 lora and Ishikei lora
Good old plagiarism, eh? Hentai artists should sue.
They should considering he draws copyrighted characters and has copycat hentai artists copying him and now a a.i. lora to imitate that style but can be mixed in with other loras
Loras themselves are plagiarism. They are also illegal since you remove artists' watermarks to create your loras. Blatant theft.
some keep the watermark and insert patreon watermarks and you have to edit them out and put them in the negatives
Any messing with watermarks is an infringing behavior in violation of DMCA Section 1202 and makes you liable to be pursued legally.
Gasp
why can't you do anything without literally stealing other people's art?
It's not stealing if nobody can claim it
the artist can indeed claim the training data
and the EU is forcing ai generators to release all copyrighted training data
which artist can claim this pic and would go though the litigation?
All artists are already going through the litigation of getting their art removed from the training data.
https://haveibeentrained.com/
but you can just save pics and train your own lora
so how does slow litigation to stop a website that will be cloned stop anyone from posting a a.i. gen pic of Tifa vs a drawn one on Cinemaphile?
>training on copyrighted images
That's the no-no. Any website that does that is illegal. You creating your own lora on copyrighted images makes you liable to be pursued legally.
So how does one go after a person who trained a powergirl lora using hentai and rule 34 artists who also don't have copyright claim to the material?
I made a powergirl lora in Shademan style, do you go after me or keep fricking his corpse?
and if you go after me do you go after all proper rule 34 artists and hentai artist who draw powergirl and have patreons and stuff but have no legal reason to draw her?
Shadman can go after you since his style includes his original stuff
but all his material is already a copyright infringement like most artists
Not at all. He doesn't just draw fanart
he would have to go though a hell of a lot of legal work to doge the Dagne Keen fiasco and all the work he did with e-girls from drifferent properties with no claim to take legal action on a anon positng on Cinemaphile, I suggest you tweet him it's going on in real time.
dude there's people who literally state
>LoRA trained on Lois van Baarles works (Loish)
that just want to literally replace the artist
that's why those models/lora are very iffy to me. if it's just the "mix" variety, i'm alright with them, since their sources are very diverse
those aren't "iffy", they are literal illegal shit explicitly trained on illegal material since Loish never said you can use her original art
action figure
Any AIChad willing to take an img2img/controlnet tackle at this? https://files.catbox.moe/ae5osq.jpg Warning: hyper breasts
>the ai can only do ATSV Spider-Gwen porn bate lips now
I wish this tech was around when the internet went insane over Lady Dimitrescue, all the generations would have looked taller and mommy-ish.
I love some Cinemaphile bait in the morning.
>>Once you've seen a few hundred AI works you start to recognize them from the thumbnail.
Has Six from Trippin The Rift ever been AI done - it seems conceptually appropriate!