Hi, regarding Post #6210373 (including parent post) this face is possibly AI generated. After checking the artist bio and posts in pixiv, they have been drawing without AI tools before but after this post Pixiv post #106626890, they have a commentary where they are now using AI tools to assist in making illustrations. For the meantime, I have added the tag AI-Assisted for the mentioned post earlier because I feel it's only the face part that is made with AI tools (although the abdomen looks kind of weird but due to the ecchi nature, this might a stylistic choice. The hair color and texture also resembles most of the popular stable diffusion models current available without any differentiation, the artist's original style doesn't match this new one.)
Please advise on how to proceed to either flag this post or be exempt due to the following clause of AI-Assisted tag: "See AI-assisted for AI-generated pictures that have parts redrawn by an actual person. Note that AI-assisted posts with sufficiently high-quality human retouching are exempt from the ban on AI-generated art." AI-Generated ruling
Hi, regarding post #6210373 (including parent post) this face is possibly AI generated. […]
That’s AI-generated with several manual changes. Not very good changes, though. The original background has been removed rather sloppily, probably with the magic wand tool, and there’s even a chunk of it left inside the scarf(?) loop around her left arm. Things like the choker look crudely redrawn. The helmet seems to be copied from elsewhere. Not sure if the bra-thing was copied or hand-drawn, but it’s obviously been added on top of the art because part of it was cut out to make room for the existing lock of hair. Anatomy from the original AI art is kinda strange, too. Just compare the direction of the pants’ waistband with how her ass cheeks are angled.
Many manual changes, but not very good ones and most of it still seems to be AI. I wouldn’t approve it.
I sorta want to upload works from this artist https://www.pixiv.net/en/users/180017 but I'm trash with identifying AI stuff. Some of the buttons look off on the dresses, and the fingers on another work are sort of janky, but I'd still like others' viewpoints first.
I sorta want to upload works from this artist https://www.pixiv.net/en/users/180017 but I'm trash with identifying AI stuff. Some of the buttons look off on the dresses, and the fingers on another work are sort of janky, but I'd still like others' viewpoints first.
The only one that really gives me pause is This one due to the weird line between the hair and right eye, but most of the mistakes on the others look like human error.
I sorta want to upload works from this artist https://www.pixiv.net/en/users/180017 but I'm trash with identifying AI stuff. Some of the buttons look off on the dresses, and the fingers on another work are sort of janky, but I'd still like others' viewpoints first.
He made this Twitter post with what appears to be a rough sketch
Man, I hate that AI has to make me second guess potentially genuine artists that do sketchy stuff because that might be how they enjoy doing things. Dang, that's depressing.
Man, I hate that AI has to make me second guess potentially genuine artists that do sketchy stuff because that might be how they enjoy doing things. Dang, that's depressing.
Don't give up brotha. You'll develop an eye for spotting robot art eventually.
The quickest learning experience is to dive head first into the bee's nest. Like this obvious A.I account:
The same over rendered effects, the same low resolution detail being upscaled, the same melting backgrounds and gibberish text, the same character always looking straight at the camera and not interacting with the scene. These generic images are uploaded over and over, that you soon realize Human Art has more flexibility and feeling going for it.
The same over rendered effects, the same low resolution detail being upscaled, the same melting backgrounds and gibberish text, the same character always looking straight at the camera and not interacting with the scene.
The images are created from simple associations with keywords, but the way we recognize images is different. The AI has no idea what things look like. It does not feel anything when arranging the pixels. It is playing a game of Mahjong with it's own connections, and showing you the end pairings as an image file. These systems are strange to observe, if you give instructions other than "make a pretty person" or "write X story" "make paintings of nature". The system sees every concept as equal, no knowledge of what is beautiful or not, what is horrible or not, that would come from us. It is not "intelligent". That is where the issue lies with trusting it. It does not understand us, or anything really. Because to understand something you need free will (to reflect onto it.) We do not know how the brain works, but it's clear that we are not our brains and they are receptors for what we really are. If you trust an artificial language or image system, you treat an iterative pattern like a mind with self-agency.
The tools available right now surely are nothing compared to ones that exist outside of public view. On imagination, when you imagine something you are not thinking of words of what you are imagining. It just appears in you. This shows the limitations of any attempt at "art" that is not from a human being - the only way an AI is "making" art is by sifting through millions of images that are shadows themselves of the inner imagination of everyone, dotted up and finding it's own patterns in them, having no idea what they look like or convey. To call any of these words or images, artificial "intelligence", is untrue.
I sorta want to upload works from this artist https://www.pixiv.net/en/users/180017 but I'm trash with identifying AI stuff. Some of the buttons look off on the dresses, and the fingers on another work are sort of janky, but I'd still like others' viewpoints first.
I think AI. twitter #1599223737261309952 messed-up ear, hair bows merging with hair, hair strands melting into each other, weird details twitter #1603035891840471040 hair melting into eyes, frilled collar disappears between the neck ribbon (which is also messed up) twitter #1604726021735190528 hair melting into eyes, all bows messed up, twintails make less sense the longer you look at them twitter #1609920625258422272 hair melting into eyes, all bows and other details messed up, row of buttons weirdly broken twitter #1612834998645854212 messed-up hair ribbons, messed-up bowtie, AI hand
I think AI. twitter #1599223737261309952 messed-up ear, hair bows merging with hair, hair strands melting into each other, weird details twitter #1603035891840471040 hair melting into eyes, frilled collar disappears between the neck ribbon (which is also messed up) twitter #1604726021735190528 hair melting into eyes, all bows messed up, twintails make less sense the longer you look at them twitter #1609920625258422272 hair melting into eyes, all bows and other details messed up, row of buttons weirdly broken twitter #1612834998645854212 messed-up hair ribbons, messed-up bowtie, AI hand
In addition to those comments, I do find the large amount of space between the eyes to be a bit unsettling.
Is it ever common in anime for eyes to be that spread apart?
I don't know which artist this is imitating in particular, but it's common for doll-inspired artists to draw wide and exaggerated faces (for example shirosaki london).
This forum thread is for checking if an image is AI-generated or not, not for any nomenclature, philosophical or ethical conundrums. Come on.
That this thread has to be made alone raises too many questions that I had to share a perspective on the elephant in the room somewhat. People are not equipped to deal with a possible super-intelligence properly, and that generative AI even got to this level is disappointing. It has no perceivable ceiling and the research is only continuing. The good thing would be if they weren't ever made in the first place.
For now we have our eyes and instinct to recognize them, and I'm glad Pixiv even lets you filter them out, but the Internet will just turn into a mess of you not being able to tell what is real and what is not in a few years, if it keeps going. It is strange how there is so much effort in short amount of time to attempt digital imitation of reality through language, images, videos, sound and speech. Prior to late 2022 the programs were nothing of note at all (like GANs which could only render somewhat-plausible cropped faces).
It's fine that this thread raises questions for you, but this belongs on a personal blog or social media. Maybe also the Discord if you want to talk to Danbooru users exclusively. The Danbooru forums are meant for functional topics about Danbooru, and off-topic discussions are not allowed.
It's fine that this thread raises questions for you, but this belongs on a personal blog or social media. Maybe also the Discord if you want to talk to Danbooru users exclusively. The Danbooru forums are meant for functional topics about Danbooru, and off-topic discussions are not allowed.
I understand. I believe there were uploads in which the user has no idea the image they uploaded was generated. There might be tools made to detect images which are not real as there is a possibility that our eyes won't be able to tell the difference once the program reaches a certain level, but there is always an "uncanny valley" that makes me know an image is fake.
There are questions that pertain to our ability to detect these images. It is uncertain for everyone, no one has really dealt with a situation like this before, and I do not have any answers. The intent of my posts were not to change the topic into a debate, but simply make aware the exponential technology that we are capable of creating and the changes it can cause in a society.
