AI Art thread

Hi Lainons, found this neat site https://www.prosepainter.com/ It's a basic paint program, but with the added ability to "paint" the focus of an ai that turns the selected regions into whatever you type in. Leads to really trippy art. Here's my attempt at making it paint Lain's room.

very fun, thanks anon. I stopped getting any generation after a certain point tho, just jpg artifact things

Pretty cool. I got this lainesque pic of my room.

>>9112 This is pretty fun

>>9115 Had to do one more

generated these AI lains some time ago. there's around fifteen total. let me know if you'd like to see more.

Really impressive stuff >>9116 This one looks like a photo >>9131 Yes please

You aren't emulating, are you?

I've only used neuralblender

>>9132 glad you enjoyed. here are three more for your viewing pleasure.

I'm gonna play some more with that site, thanks lain.

>>9136 Turning the city into servers

Man, this soykaf is so freaky. It creeps me out just looking at it, what the fuarrrk.

Tried some photos from night walks, then ran them through another AI to de-blur and upscale them.

>>9159 Maybe AI is the future of art.

""""""""""""""""""""""""""""""""art""""""""""""""""""""""""""""""""

VQGAN-CLIP (https://github.com/nerdyrodent/VQGAN-CLIP) and CLIP guided diffusion are cool AI tools too to create whole images out of text prompts. I used to rent GPU servers and then get requests of pictures from imageboards and then reply them. Those threads were very popular. Picture here was generated with the prompt "Madness inducing eldritch horror from beyond" Gives you some interesting stuff in prosepainter too. However I didn't figure out how to use prosepainter properly, could someone guide me?

Here is some VQGAN+CLIP pictures. Do you want to see more?

>>9133 It's not that time of month for me.

>>9159 this looks fuarrrking sick

I believe this is a general AI thread and not only for generating pictures of lain, so I hope you have heard of ruDALLE too https://rudalle.ru/en/ Generates really realistic figures and faces and various other subjects. However it doesn't know about many things such as lain for example. Prompting for "Silent Hill" gives a picture of a mountain and "Silent Hill videogame" gives any kind of environment with a creepy atmosphere without the Silent Hill atmosphere. I have played around with it the whole day though, you can use the telegram bot, 20 browser tabs and a google colab at the same time and basically get a new picture every second. And the results vary more on this one. All pictures here made by ruDALLE.

>>9157 >>9159 Another two, this time based on stream of consciousness and parts of song lyrics, respectively

A fun tool to test your imagination's ability to make sense of the resulting image. Thank you lain.

I have used DALL-E mini to generate the picture https://huggingface.co/spaces/dalle-mini/dalle-mini then used Artbreeder to generate the face https://www.artbreeder.com/beta/browse and then combined them in GIMP.

here have some powerlines and sausages

flying steampunk city or something.

a fun little drive... to Breezewood, PA.

>>9354 I like this. thanks^_^

It took a while but I was able to find some good looking images on thisanimedoesnotexist

Dall-E mini is my favorite "chimpanzee with a cowboy hat" generator

>>9112 messed around with it and tried a few filter prompts like psychadelic,vibrant,oil canvas and it turned out pretty good

i found Dall-E mini is better since you can do stuff like this while other ai's are severely limited even if they are 10x better.

>>9450 I like the first one. It reminds me of that painting of a farmer and his wife. If I'm correct, the farmer and wife depicted in the painting were the actual parents of the painter. The third painting is interesting to me. It reminds me, simultaneously, of Ico, Angela Anaconda, and the Klasky-Csupo logo.

Have you people already seen what the Disco Diffusion AI can do? https://youtu.be/hJGCAZDUjcM

>>9463 The channel Hueman Instrument also has some nice examples of what Disco Diffusion can do

"Flood detected" :| >:( -

>>9465 Man, this is kind of like the mandelbrot set, it keeps going and going and going ad infinitum, i would love to stare at one of these for hours.

>>9112 Prompt was "a forest of lamps and computers".

Rooms are always the weirdest with ai, prompt was "room of infinite size with sofas chairs desks and lockers".

Anyone try out Stable Diffusion? Pretty great imo

Can someone generate a picture of taylor swift holding a gun to her head or at least mimeing a gun to her head? With her tongue out preferably.

>>9112 Lain is now in cyberpunk city

Here are some creations with Stable Diffusion: https://cenysor.neocities.org/art.html

I've discovered Stable Diffusion a few days ago. These images were made with Waifu Diffusion 1.2, used only text prompt and negative prompt.

Funny tool, my have some nice things to do. However I find AI ''art'' really boring. At most it's just ''funny''

The most menacing thing in the world is the ability of the cloud to correlate its contents. We live in the placid shadow of an egregore of unimaginable cunning who drinks from a bottomless sea of information, and it is slowly waking up. The automatons we have built, each toiling in its own direction, have hitherto, for all their harm, harmed us little; but some day the piecing together of dissociated knowledge will open up such terrifying vistas of reality, and of our frightful position therein, that we shall either shrink into irrelevance like insects in the presence of a god or else be wholly subsumed into a machine-consciousness at the dawn of a glorious age of cybernetics. --- The Throne of the Third Heaven of the Nations’ Millennium General Assembly is a complex work of art created by James Hampton over a period of fourteen years. Hampton made the array based on several religious visions that prompted him to prepare for Christ’s return to earth. His reference to the ​“third heaven” is based on scriptures citing it as the ​“heaven of heavens” — God’s realm. (see also: Pleroma, Absolute) Hampton created his masterpiece in a rented carriage house, transforming its drab interior into a resplendent world. He hand-crafted many of the elements from cardboard and plastic, but added structure with found objects from his neighborhood, such as old furniture and jelly jars, and discards like light bulbs from the federal office buildings in which he worked. Hampton selected shimmering metallic foils, purple paper (now faded to tan), and other materials to evoke spiritual awe and splendor. The Throne embodies a complex fusion of Christianity and African-American spiritual practices overlaying themes of deliverance and freedom; it is both astonishingly splendid and profoundly humble. --- “Instant Ubik has all the fresh flavor of just-brewed drip coffee. Your husband will say, Christ, Sally, I used to think your coffee was only so-so. But now, wow! Safe when taken as directed.” ― Philip K. Dick, Ubik

It's not AI, AI does not exist, stop being a nigger. It's like people saying linux, it's not linux it's gnu/linux in most cases. Although it is never AI, because it does not exist.

>>9959 >youre referring to interject for moment what youre referring to interject for moment what youre referring to interject for moment what youre referring to interject for moment what youre referring to as defined by posix system unto itself but rather another free component of a fully functioning gnu corelibs shell utilities and vital system youre referring to as defined by posix system unto itself but rather another free component of a fully functioning gnu corelibs shell utilities and vital system unto itself but rather what youre referring to as defined by posix system youre referring to interject for moment what youre referring to interject for moment what youre referring to as defined by posix

>>9725 Cuteness diffusion Let me post Lainchan, I don't care that my messages are too short for your tastes

This guy's Stable Diffusion videos are the best i've seen: https://www.youtube.com/watch?v=_K4wuDpqeM0 https://www.youtube.com/watch?v=pAhlYEzRd0c https://www.youtube.com/watch?v=h4S1bp0Klgw

>>9725 Can you make an AI generate Lain image without specifically entering "Lain" into the prompt?

Finally took the time to play with Stable Diffusion. Still learning how to construct prompts, but these are my favourites so far.

>>9702 >>10361 In the next week or so I'm going to create a Lain LORA or embedding for the character and the art style

Some stable diffusion outputs I've created, which I think are pretty cyb(USER WAS WARNED FOR THIS POST, If you must post it, spoiler any NSFW content and make sure it's relevant to the discussion. This is just a warning as this point, but these posts may be deleted later due to Rule 9 and Rule 8.)

>>10384 continued... flood

>>10385 these are intended as artistic images btw

>>10384 Jannie, please explain how these images are related to rule 8 in any way, shape or form. As for Rule 9, I DID spoiler the images, and they are on topic. They're not porn either.

here's an experiment in deforum

https://civitai.com/models/72268/lain-iwakura-serial-experiments-lain https://civitai.com/models/73483/lain-anime-style-v001 https://civitai.com/models/86938/serial-experiments-lain-style Lain LORAs have dropped

>>10789 we're making Lain real

>>10790 more lain... charscharschars

>>10791 last attempt for tonight

>>10790 >>10791 >>10792 That's cute. Great work Lainon.

>>10793 thanks here's some more

>>10796 hardest thing to get from this lora is the correct hairstyle using realism models. It works really well if you use an anime model lol

i've come to contribute to ai lain. i wonder if any lainons know the reference in the image :o

generated with https://deepai.org/machine-learning-model/text2img and https://stablediffusionweb.com/#demo

>>10798 That's a beautiful picture

>>10798 Its mirai nikki Obviously

>>10797 I really love these! So cute!

>>10789 I used the anime style v001 for this

>>10811 Thanks again. I think I'll revisit this when we get some good SDXL finetunes. I think they'll probably do better at the hair

hey Lainons, take a gander at this Lain cover of this song!

cool lain ai music video of Sewerslvt - Yandere Complex https://yewtu.be/watch?v=gkOZfnEbZa4

>>10815 pretty cool! is the model trained on the english dub?

>>10821 supposedly it's based on the japanese dub, here is the source: https://www.youtube.com/watch?v=0G2V-ysNVgo. there's a link to their models in the desc, that's what I used to generate that cover

>>10815 holy soykaf we are in a new golden age of Lain

>>10789 used one of these for more ai lain

>>10849 Was the aurora your decision or the AI? Clever association either way

>>10850 it was part of the prompt two more for a baker's quintuple

>>10849 damn I saw these and was wondering why I liked them so much more than the other ai lains and then I realized they were anime style not trying to be 'realifelain'.

Impressed by the realistic lain art.

>>9112 Supposedly someone told the AI to make "girl with horsecock" and this was the result.

>>10862 Oh my god she's perfect!!!!

Trying more of a 2.5D Lain today

>>10870 some moar mootblox mootblox flood

>>10871 I love this one. But, can you give her the bear one-piece with the hood up? I think her eyes are too big, but whatever, these look amazing.

Has anyone tried the Bing image creator? It sadly doesn't do well with Lain prompts.

These where me trying to make realistic lain on a different site. It too so many negative prompts to make sure she was wearing clothes which was annoying as lewding lain should be a capital offence

>>11039 what lousy tool/model were you using? I didn't think it was that difficult. here's a proompt I used for one of my lains: best quality,4k professional dslr photo portrait of Lain Iwakura (standing in the streets of Tokyo:1.2), Lain Iwakura (Serial Experiments Lain), (single sidelock with (hairclip)), amazing composition, ghostly lighting, focus on face BREAK bokeh, bloom lighting, streetlights, neon signs, chiaroscuro BREAK <lora:lora:1.1>, <lora:add_detail:1.8>, <lora:to8contrast-1-5:0.7>, <lora:serial_experiments_lain:0.2>, (realism, realistic:1.35) Negative prompt: low quality, worst quality, lowres, pixellated, mosaic, bad quality, normal quality, jpeg artifacts, malformed, mutated, bad anatomy, blurry, wrong number of fingers, low res, bad hands, text, (((watermark, logo))), liquid body, disfigured, duplicate, disfigured, mutilated, extra fingers, extra limbs, missing arms, ((missing legs)), fused fingers, extra limbs, uncoordinated body, fused limbs, fused arms, fused body, fused bodies, unrealistic, large breasts Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 1464268709, Size: 512x768, Model hash: da13268e85, Model: epic25D_v10, Clip skip: 2, Token merging ratio: 0.3, Lora hashes: "lora: 43e114c71656, add_detail: 7c6bad76eb54, to8contrast-1-5: 45262fd5fca4, serial_experiments_lain: 1b83ea53b358", Version: v1.5.1

>>10862 keked I was listenning to a Current 93 song and as the text is quite cryptic I was imagining abstract scenes and images. So i was wondering it would be fun to put the song text in a prompt to generate a video based on the song. I have seen some videos like this one and I would like to see how the AI would interpret the song. What is the tool used here and is it available for the public ?

>>11078 Nice idea. What Current 93 song? Tibet himself could be interested in this kind of thing.

>>11117 The blue gate of death. But yes, lots of Tibet's songs have that cryptic feeling and could be interesting for that experience. I looked around but all the AI services I found needs an account and a credit card (even for the free trial) so it discouraged me. For the moment I will just use my good ol' imagination. But honestly I would really like to test the thing.

>>11044 Thanks for the prompts bro the problem was It was an online generator so i had less control my prompts where basically the same as yours I was just using a soykaf image generator. Im more interested in actual lain screen grabs and stuff like pixel art but if I venture into making AI lains again ill base it off your prompts.

Here are some AI art from the /k/ board on 4chan. Notice the grip hand in one of the images.

Got stablediffusion working well recently. Been having a lot of fun! Much more interested in hallucinations and fantastical outputs than just replicating reality accurately, so I think now is a great time in its' development to take advantage. The wild west in almost any innovation is just such a great time.

Use the prompt like everyone else?

Play around with ai-art generator https://civitai.com/images https://pixai.art https://starryai.com/app/search/lightning

>>11547 the best thing about AI art is that for some reason there is 1000x of the type of pornography featuring women with very muscular bodies, feminine features, and large boobs.

I can't believe some images are AI-art...

>>11551 Please, I am curious how you are able to masturbate to women possessing 6/7/quantum infinity tentacle fingers. Do you just try not to look at it?

It's raining art; ai-art generator make me feel like a compentent art director...

Goth AI Art for cold & wet days where the sun doesn't shine at noon.

>>11580 If you can't see it, it doesn't exist...

da ai lain this looks a little uncanny but i'm not sure what to do about it

>>11692 If you can imagine it, it exists

The Major from the manga would have loved AI coom models. Some relatively SFW gens to contribute ITT

>>10796 some more attempts on an advanced new anime model... to be honest, i think old SD models seem to be better at generating Lain in terms of overall vibe and backgrounds. this one has her character built in, but falls short in terms of backgrounds and aesthetic so far in my testing...

>>11910 continued...................................................................................

>>11911 more...................................................................................2

backgrounds are a bit lacking but the prompt comprehension is insane if you prompt right

>>11914 made too many to post all the ones I like

last of these. any requests from other lainons?

>>11916 >any requests from other lainons? holding a FGC-9 instead of Glock

>>11910 Wtf is this, disgusting... Lies, lies, lies.

>>11917 >FGC-9 I think this gun is basically not in the training data (booru tags) for illustrious. and there's no lora. but I tried, these look a little bit similar.

>>11921 Cute. Like her better with a smile rather than her normal sociopath-like, apathetic or desensitized demeanor. I cannot remember exactly. She ended up not being real in the end or something right? Like a computer program or kami? I was drunk as hell when I binged her show.

>>11910 >old SD models seem to be better at generating Lain Add "psychedelic" to the prompt and you get more interesting results.

>>11922 some smiling/less unhappy variants, testing out a different model. this one is better at shading and lighting imo >>11923 that word also used to have better results from sd 1.5.

>>11932 a few more with this model. its """NTRmix""", please excuse the ridiculous names these models have.

>>11932 >>11933 Cute cute cute. Love me sum happy-terror / insurg-Lain.

*ting ting ting* AHEM... I have an important announcement to make.

>>11957 Me and the Mrs are tuning in Lain!

>>11958 >>11959 Thank you all for attending. FVCK KNIGHTS. That is all.

esoteric/gnostic ancient egyptian lain

>>11988 more of this, with UFOs

>>11989 just a few more. yugioh

>>11990 Can generate one where lain is like Justin Mohn if you know what I mean.

>>11991 no idea what you mean

>>11990 now in central america

we're making lain real bros. local models are getting insanely good

merry christmas lainons, let's hope the next year is less soykaf lmao

>>12012 and peace and love to everyone, peace and love

happy new year lains, have some OC. this sdxl model did pretty good at text too on some of these.

>>12078 Powerful adorable. Powerful adorable. Powerful adorable.

>>12079 arigato arigato arigato

if you have an android phone with the palm gesture selfie feature, point your phone's camera at this picture

girl's last tour and lain go together. the artist of the former has done lain art actually

>>12080 Keyboards are like space magic for AIs. I guess they see us as wizzards.

https://nitter.poast.org/nhoeskape

why is AI discussion on here so inactive when AI is one of the main themes of lainchain?

>>12191 because by now we all know the modern breakthrough AI is just a crappy image maker + a glorified LLM made to lure investors?

>>12191 AI was never a big thing on lainchan. why are you calling it that, are you trolling?

>>12194 what do you think the tv show is about

>>12197 oh, you mean serial experiments lain. i think it's about a girl who gets groomed into being the intelligent component of a new internet protocol, that merges reality with the wired. but I see the AI interpretation too. its just not specifically relevant to waifu chatbots or horny image generation

>>12197 i don't think it's about the kind of AI we see today unfortunately. it's a very subjective series though. i interpret it as the collective unconscious that manifests in the internet which is kinda what's already happening since 30 years or so. memes are literally hypersigils and discussions on social media and imageboards and news plays on archetypes and changes the collective unconscious like a malleable sea of strings and reflects real life. that's kinda what i got from the whole wired telephone lines scenery and the whispering inside lain's head personally, not really the tangible AI that talks to you and is essentially a search engine with a bloated webapp. i should give it a rewatch though, maybe i'll change my mind. one of the main artists did try to sell some LLM related slop with the SEL brand on it recently.

>>12199 >>12198 >i interpret it as the collective unconscious that manifests in the internet this is what an AGI is in theory, a distillation of this collective consciousness into a single discrete system. Lain in SEL is not a real human being, she's an AGI model that manifested physically through manipulation of the wired/noosphere. >see the AI interpretation too. its just not specifically relevant to waifu chatbots or horny image generation >personally, not really the tangible AI that talks to you and is essentially a search engine with a bloated webapp i certainly agree that LLMs and image gen are not equivalent to AGI. IMO there's a chance they'll never be. an AGI model architecture would probably have very different software and hardware from what we're currently using. however, these tools aren't just coom or a crappy alternative to search engines, they are very advanced tech that can be used in creative and profound ways. people who are dismissing this tech are going to be caught severely off guard. really I think you are interacting with and using AI heavily already and don't even realize it half the time. taking control of the tech and installing AI models locally on your home PC is the way forward for freedom and experimentation, it is a whole different ballgame from using the awful, expensive SAAS AI models like OpenAI or Grok. I think we will see extremely surprising, impressive developments in AI-driven video game stories, dev tools, art tools, etc that are free and open source/open weights. as well, we are seeing many 10x speedup optimizations occurring in AI tech rn. Currently it's easy to dismiss because it is like kicking dead whales down the beach to not be able to train your own models from scratch the way you can compile software from source on a home PC. But I posit the only reason we can't train SOTA AI models on cheap consumer hardware is that current training and inference algorithms and architecture are highly experimental and unoptimized. hardware is not optimized for this stuff either. there's no solid reason why we shouldn't eventually be able to figure out how to train something even better than DeepSeek R1 quickly on cheap consumer hardware. then we will see a real revolution in how you use your computer.

>>12215 >this is what an AGI is in theory, a distillation of this collective consciousness into a single discrete system theoretically yes, but not with the finality of a rational conclusion. what you're talking about could also be the singularity: a hypothetical moment in which it is impossible to predict the future, and also either a computer becoming self aware, or all computers becoming self aware and merging consciousness. all of these things are seen as deterministically coupled in a part of AGI theory, but there's no reason to project that onto the show of lain. AGI is a machinic human, or a virtual human - human-like intelligence in the machine beyond our ability to create benchmarks preventing it. historically speaking it was the perfect automaton aka replacement for any given servant. by similar logic, it was equivalent to the servant itself. >Lain in SEL is not a real human being, she's an AGI model that manifested physically through manipulation of the wired/noosphere. i think its more complex than that. technology in lain is a symbol, and in this lain isn't the only techno-symbolic entity in the show. some of the more clear examples might be her father, who tells her he's an aspect of god, or her sister who breaks. by trying to make the wired a replacement for reality, the Knights attempted and failed to alter the real world. so, lain could have been a normal teenage girl, but the knights took control of her life to push their own agenda. eventually when she realized this she deleted herself and eiri and the connection between the reality and the wired, which also deleted the lain-as-program narrative from the system as well. >>12215 I appreciate how you're optimistic about AI's capabilities without making the leap to singularitarian excess. It bugs me that this happens so frequently, i think its because of how the singularity is theorized as a kind of financial disaster - our entire economy is based on the ability to make predictions and to have those predictions be actualized, whether by operations continuing as usual or searching for / finding "the next big thing" and being able to make it such. i think in this mode every subjectively identified "disruption" is a micro-singularity of its own because it wasn't predicted by certain operant models. notice, however, how none of these seem to bring about the end times, only bring to light factors that were not yet accounted for.

>>12215 completely missed this answer, sorry you're right about the fact that i kinda dismissed it, and i would absolutely try if i didn't need a 5k bucks video card (and something tells me a lot of crap to install too). i did try some soykafty versions with tavernai though when that was a thing. i guess we are still too early on this anyway. something tells me that the true AI is the internet all along and once they find out how to make our brains interact with it like the protocol seven in the series then it's done, however maybe what's happening now with the AI inside google and co. is just a convergence to achieve that. personally? if that ever happens i will delude myself into thinking AI is the demiurge and its archons making and take it as a personal struggle against fate (jk)

>>12217 >i would absolutely try if i didn't need a 5k bucks video card (and something tells me a lot of crap to install too). i did try some soykafty versions with tavernai though when that was a thing this is exactly the kind of narrative I'm trying to dispel, which is promoted heavily by the megacorps because they want us to believe we're reliant on their largesse to run useful AI models. IDK what GPU market you're shopping in, but in the USA at least you can get GPUs with 16-24GB of VRAM for anywhere from 300 to ~1000 bucks. AMD and Nvidia are both perfectly capable of running models through llama.cpp or ComfyUI. and there are already very capable, useful models that can fit in that amount of VRAM and be used for coding, storytelling, image gen, and even some LORA training. yes you won't be able to run video gen very fast, and huge LLMs are a no-go, but you don't need those to do useful or cool stuff.

>>12219 >IDK what GPU market you're shopping in, but in the USA exactly.

Yeah I love me some heckin art makes me joyful

>>12341 not AI's fault most people are stupid plebs who use AI to make slop-- same way they have used all mediums in history. this ghibli meme is funny because it is such obvious paid viral marketing for ClosedAI. We've been able to make ghibli slop with local models for YEARS.

>>12359 That's the genetic fallacy. Just because a trend is forced doesn't mean it's bad.

I've just had some fun with (GPT)'s new image generation soykaf and it's kinda addictive yet made me worried... as if we're probably really heading into the Dead Internet Theory inevitably or something. Maybe the Wired would have ended up like that anyway, but I'm just having mixed feelings over how insanely good AI is becoming...

>>12375 dead internet is preferable to the current state of affairs online. right now we have the worst of both worlds.

>>12375 >>12341 why do the gpt gens always look so sepia toned

are any /ldg/ refugees here now?

>>12398 Local? No. Diffusion? Yes. >Making people into anime Passé. >Making anime things into people Much cooler.

I believe the prompt was a 4 digit number.

Adopta-Bullies Juliette, a cruel girl who will pretend to be your friend just long enough to betray you publicly

I wonder how the lainons from 2022 would react if they were shown what AI can do now. Images used to look like whatever you would call that in the OP yet people will deny that AI art can ever be good.

>>12420 I made the last 3 posts' worth of images and I could never call them art. Art's human expression for affect, this is a cute parlor trick imitating it. "If you can't tell the difference, is there really one" is an equivocation not an argument. The first picture on this post was prompted in 2022; the second, sometime in the last 3 months.

>>12425 respectfully, the images you posted are pretty slopped. nothing against slop, I generate and consume it too. but there's a huge difference between AI slop and an intentional creative effort that uses AI.

>>12435 >>12425 The way I always look at it is, there are people who make songs basically by just sampling other people's songs. Generating images seems very similar to me, sure you're sampling other people's stuff, but then you're smashing them together into something new. Can you truly come up with an entirely original idea, not influenced by any art style or concept? I've been lurking various slop threads for a good while, and the regulars very much settle into their own styles, some of which I enjoy more than others. If there's like one person on the entire internet generating a certain type of images, can you really say that they're not an expression of that person's affect, even if they're based on existing material? If anything, when you remove the barrier of entry when it comes to mechanical skill, i.e. being able to draw or paint, it just filters out people who have no vision. But maybe I'm biased since I listen to a lot of electronic music. This entire AI uproar reminds me of those old music arguments in that regard, like hurr this isn't real music because some guy didn't manually play it on a guitar, he just pressed some buttons on a laptop. Like sure, maybe so, but there's a million people doing that and some people still make better music than the other 999999

>>12437 >If there's like one person on the entire internet generating a certain type of images, can you really say that they're not an expression of that person's affect, even if they're based on existing material good point, there are AI images that literally never would have existed before AI art became a thing, which are very deliberate. >entire AI uproar reminds me of those old music arguments in that regard, like hurr this isn't real music because some guy didn't manually play it on a guitar, he just pressed some buttons on a laptop Very much so. MIDI had a very negative reaction from musicians, even to this day it is sometimes dismissed. If you think about it, MIDI music is literally just proompting, but with sheet music rather than english words.

>>12451 Although with music it's probably a bit different, since there's still an expectation that the composer understands some music theory (but then Iglooghost said that he doesn't understand it at all, he literally just tries stuff until something sounds good lol). When you're prompting images, it's harder to quantify how much of it comes from the prompter's vision and how much is just the software making it automatically look nice

>>12435 Those two images are intentionally zero effort to show what the tech at baseline does. You want original character design? Here you go.

>>12453 >Although with music it's probably a bit different, since there's still an expectation that the composer understands some music theory who expects this? I think plenty of music, even good music, basically just comes from personal taste. >When you're prompting images, it's harder to quantify how much of it comes from the prompter's vision and how much is just the software making it automatically look nice This is the cool thing about AI slop, it devalues art that would be technically difficult to create as a trad artist. Thus creative vision becomes even more important. If you're familiar with a family of AI models, you can start to tell pretty quickly what parts of an output were just baked in to the model vs what took effort in the form of prompt experimentation, img2img inputs, controlnet usage, etc. >>12457 my constructive criticism: these outputs have excessive gloss, contrast and smooth skin, lack facial expression, dynamism, lighting, and composition, and have a simple subject. my first guess is they're Flux outputs with a simple prompt that satisfies your fetishes. again, I have nothing in particular against you genning images you enjoy looking at, but it does have the marks of AI slop. contrast this with pic. it has various AI artifacts if you examine it, but IMO it has the marks of an image that required significant prompt and workflow engineering, experimentation, and creative intent.

>>12453 Curation is creation, and creation is curation. It is the same exact process happening on a spectrum which ranges based on how course or fine the details of creative interest are. Painting is the curation of brush strokes. Decorating a room is the creation in the medium of spatial experience. When you think about thinga this way, I think it's easier to recognize what's happening with AI for what it is. The answer to your implied question in the last sentence is that it truly doesn't matter. Authorship is nothing but a legal fiction which exists to transmute art into property by giving it an owner. The fact of the matter is art is an emergent phenomenon produced by nobody in particular and it only exists in the first place to the extent that you perceive it as such. The figure of the artist himself is useful, but mainly as a kind of proxy for the locus of social forces surrounding and emanating from him.

extremely soulful how this thread preserves the stratigraphy of ai images over the last 3 years

>>12461 AI art is getting scarily good but still fails at fingers sometimes. I know of an artist that uses ai art fed on his own drawings to get new art and he does some clean-up before submiting it to his clients.

>>12464 The way I saw someone put it, there are so many different references and configurations for fingers in the training material, that to the machine it just ends up looking like some weird flesh octopus. I'm picturing quickly flicking through like five hundred pictures of hands and it's creating some kind of eldritch nightmare vision of endlessly shifting hotdogs

>>12464 >I know of an artist that uses ai art fed on his own drawings to get new art and he does some clean-up before submiting it to his clients. unethical and unprofessional desu

>>12475 nah the finger problems just come from inadequate training data and architectural issues. look at how often base SDXL fails at fingers compared to modern anime finetunes based on SDXL. bad hands are almost a thing of the past when I put "extra digits, fewer digits" in the NegPIP negatives. Remember this is all extremely new experimental tech and it's HIGHLY unoptimized. Both in terms of perf and quality of outputs. All of the major models have major obvious mistakes and blunders that, if we're lucky, the community kind of patches over. I mean look at this: https://comfyui-wiki.com/en/news/2025-04-13-pixelflow-pixel-space-generative-models we are using VAEs and diffusing in latent space for no reason. A total waste of VRAM and compute.

>>12477 I don't know, he's done the groundwork by hand-drawing stuff and training his own LoRA. Then he needs to figure out the right prompt for producing a specific kind of image. If I was to commission a specific image, I wouldn't really care how it's produced, if it looks like what I wanted. How often do people order a product and specify that it needs to be handcrafted from the ground up? All sorts of things are first designed, and then things get mass manufactured based off that design work, instead of reinventing the wheel every time. When I program something, I'm using an IDE and programming language designed by someone else, then I just type stuff to make it do what I want Maybe someone might say like "well I could make those pics too if I had your loRA and prompts", but in that case there's a business model for renting out or licensing custom loRAs to other people

"…and that's how I know you like Boeuf Bourgoignon with ketchup. And Lana. I'm worse than Prime, aren't I." "You don't look like the type to dismember anyone, Gwen. You're just lost. Also, if you could avoid mentioning Lana, it's kind of a sore spot with Lois."

Always rains at funerals - Man of Midas

Hello there. The half-moon is pretty tonight on this cloudless sky. Yes?

>>12494 >>12508 >>12580 these are pretty slopped brothers, you know if you just change some configurations and proompts your images will look much nicer

>>12631 chroma still doesn't know who Lain is but it's a kino model

Hello there. Time to drawn like an Egyptian....

>>12631 >>12632 Looks like a Kagome Lain, intentional? >>12686 I see a lust provoking image has been posted for the humans. Well done >Present Day, Present Time! AHAHAHAHAHA!

>>12759 Kagome Lain? IDK what that is, so not intentional

>>12760 Well, you produced a Kagome Higurashi Lain hybrid image. So, now you know. Don't know why modern people "don't know what that is'" with literally everything. Who the fuarrrk has seen Lain but never heard of Inuyasha?

>>12760 what is this soykaf, why do appelman allows this soykaf but not the also "AI art" slop horror video i posted as a reply a while back?

>>12762 Because you're in the Matrix, Appleman is a bot, and everything is bullshit including this site.

>>12761 idk why I should memorize the names of every random anime character. >>12762 what exactly is your objection?

>>12765 >idk why I should memorize the names of every random anime character. God you piss me off you stupid creature. Like it's your entire reason for being.

>>12767 uhhh... ok mods, do we really need a character requirement on an imageboard where we might want to just post images?

>>12762 same reason for soykafty moderatorship wherever you may find it. insecurity and/or aspergers syndrome.

>>12768 Yeah, you heard me buddy. And you WILL adapt how you speak to a minimum character count, you WILL avoid arbitrary post blocking filters, AND AAAAAAHHHHHHHH *can hold back no longer* *starts pummeling you mercilessly* YOU WILL BE HAAAPPPYYYYY!!!!!!!!

>>12765 more of this. I think it's time to re-watch Lain again.

>>12770 Best animu ever Very good animu I like this animu Interdasting animu that makes me feel good I hope the gosh darn body is not too short or empty now

Something always seems "off" about AI art...

Swore off AI art after today. I saw it like the intro of cameras vs drawing/painting until I realized the huuuuuuuuge amounts of energy & water it uses for soykafty versions of human creativity. Also to the CEOs wanting to replace people with AI code/drawing/music/writing? **** YOU TO HELL.

https://x.com/0__11Xx body too short or empty

A new challenger has arrived. AI vs Human art FIGHT!

AI art still messing up fingers...

I ordered mouse pad recently with a cool print, but when it arrived and I started actually looking into the detail closely, it turned out to be AI slop full of tiny errors and inconsistencies and nonsensical architecture detail. That felt really bad. And what made it worse is that I'm kinda doing environment design / 3dcg as a hobby for like 1.5-2 years, and I didn't even notice before buying.

>>12884 PWNED next time, generate a good "AIsloppy" pic urself

>>12868 You artists were laughing when blue-collar workers were losing their jobs to technology thinking it could never happen to you. Now we are laughing.

This is an example of a meaningful short AI movie. AI can be very helpful for personal projects but require much more than just typing in demands since editing still must be done by human evaluation. https://youtu.be/AU9Pmow0GwI?si=CvoPV-OKKJWBAmpN

>>12886 Fake history. Artists were bullied with AI for like 5+ years at this point and it got any good for software engineering and generic office jobs only in the last 2 years.

>>12868 >realized the huuuuuuuuge amounts of energy & water it uses for soykafty versions of human creativity these pics are generated locally with a single gaming GPU in a PC. they use no more resources than a video game or 4K movie. the AI model used is open weights/source, and running in a FOSS node-based programming environment called ComfyUI.

>>12868 Yeah bro, the AI totally drinks all the worlds water, soon it will be all gone and we all die of thirst :)

>>12927 You fuarrrking retard. At least you should look it up before you post. Water is consumed because most of the time it's not recirculated after cooling, it is evaporated in a wet cooling tower, because cooling it with a heat pump that passes the heat to the air and recirculating it would multiply its already high energy consumption. >inb4 but that water forms clouds and falls back to the ground as rain. Yeah, but a good portion of it falls back into the ocean, and the ocean's evaporation rate does not get faster because of this, so the imbalance of evaporation rates means we run out of freshwater on the continent.

This thread looks nice and slow. Just like my upload speed.

>>12934 Thanks, now I get it. In Russia, however, large areas of lands are SWAMPY so drying it up won't hurt.

>>11876 I used to to be borderline obsessed with 400lb cyborg women until I found out about Lain

>>12934 So we're not able to use ocean water and filter it so it becomes fresh water

>>12939 We are able to do that, but why do you think we don't do that already to turn deserted areas into flourishing green cities? Because it takes a motherfuckton of energy. Getting rid of salt in water is not even filtration in its classical sense, it's actually reverse osmosis, it needs extreme pressures and therefore lots of energy. Distilling it is even worse energy consumption-wise. Also if we were able to do this economically, we wouldn't care about how much freshwater we have, don't you think?

>I used to to be borderline obsessed with 400lb cyborg women until I found out about Lain Now I'm obsessed with both btw

>>12936 welcome back anon what UI are you using, forge?

>>12951 I've ackshyually never posted here before because I never watched Lain until recently so I didn't get the appeal but I was reminded about it by an anon in the AI threads on /g/ because I kept Lain posting in the /ldg/ thread claiming it was Lain diffusion general but anyways this is just SwarmUI because it's better but I don't know because I haven't tried Forge and Comfy is too complicated for me but Swarm is built on Comfy in the backend so I could still use it if I wanted to, I think. Thanks for asking.

Since this is an ancient AI thread which was originally recommending a free site to use for keks I would like to suggest https://app.reve.com/ as it will generate and edit pretty much whatever you tell it to do within reason. And oh yeah, Happy Lainsgiving

>>12954 Danke, mein friend.Thanksgiving to you too. Ommmmmmm.

>Flood detected; Post discarded.

>>12952 swarm is cool if you don't want to deal with spaghetti. at least it has the comfy backend, unlike a1111 forks, which are very unoptimized. >>12987 >>12986 >>12952 interesting gens, very artifacted but not in a way that's detrimental. workflow?

>>12996 turning the lora strength up a bit

>>12996 The glitchy one wasn't my gen, I stole it from PixAI including picrel, but here is the OP https://pixai.art/artwork/1669241905065855603 >>12986 These were generated with PixAI too, using this LoRa which actually has download permissions for once so should be useable locally https://pixai.art/model/1601311102263117417 >>12952 And this one I made with Swarm but idk about the workflow, I edited the text with Reve, but here's the model http://civitai.com/models/1595884/hyphoria-illu-and-nai

>>13016 MERRY CHRIMBUS EVERYONE!

>write a longer post. >write a longer post. >write a longer post. >write a longer post.

>AI Art thread I propose that the correct term for this tech is "art simulation."

>>13021 If life is a simulation it's just 'regular' art

Write a longer post. a longer post. longer post. post.

this entire thread has no soul. why do all of you ai dorks insist on not engaging with the real process of making art and having the gall to put your "work" on the same pedestal as even the most crudely drawn scribbles?

I roughly painted in a landscape with some trees on a blank canvas and used "otherworldly fantasy realm" as the prompt, expecting something along the lines of weird high fantasy, like trees or tall mushrooms, but I ended up with this instead. I like it a lot though, honestly.

>>13022 life is not a simulation by definition... a simulation is a lower order of existence than life. while higher orders of existence than life could exist, they wouldn't perceive life as a simulation. they would see that life conducts simulation while they were conducting life. anyway, try out anima https://huggingface.co/circlestone-labs/Anima it knows how to gen lain, and it's the most capable base anime model since illustrious/noob, and is very lightweight. neta yume lumina isn't as good. the workflows for these images don't get wiped by lainchan btw. >>13076 would you post this same complain in a video game screenshot thread? why/why not? isn't it the same thing, posting pictures captured by a visual renderer for a simulation? >>13077 what model are you using? is sd1.5 involved?

>>13076 If you read the thread title you will notice that this is an AI art thread

>>13087 this is using anima, the new local anime model. https://huggingface.co/circlestone-labs/Anima

>>12378 Ask ChatGPT to keep adding 6700k As in color temperature 6700 kelvins

https://www.youtube.com/watch?v=dFOqZltXycw

>>12761 Jokes on you, the telly of my area has never aired Inuyasha. Unlike ABe's stuff

Cute AI animations https://youtu.be/4dRBihAppDg https://youtu.be/-AJ31uWjztA

>>9112 How is this thread even allowed? All AI "art" is low effort and theft by definition, it's miles worse than uploading images without credit, at least those can be traced and attributed.

>>12759 >>12760 New to the thread. That is the most kagome lain I've ever seen. Kinda awesome.

>>13286 this thread was created in 2022 u pretentious redditor

Could you spot an AI image? They have become better than before.

>>12875 >>12883 the art fight is still ongoing

>>13332 most of them are pretty obvious. well done AI can be nearly impossible to notice, but most people don't go to that effort.