AI Art Seething General

It's a bit funny and a bit sad to see social media users realize (about 15 years too late) that the companies running these websites harvest and sell their data without concern for the individual. But, I mean, where did you think these models came from in the first place? They didn't just fall out of the sky. Trying to poison the data is too little, too late to have any real effect now. Big Data is in full swing, and these latest deals are just companies realizing they have more to gain by charging other corps for direct access rather than letting them scrape or scrape-by-proxy.
 
  • Agree
  • Like
Reactions: ADHD Mate, N Space, Ether Being and 9 others
Mr.Miyagi said:
It's a bit funny and a bit sad to see social media users realize (about 15 years too late) that the companies running these websites harvest and sell their data without concern for the individual. But, I mean, where did you think these models came from in the first place? They didn't just fall out of the sky. Trying to poison the data is too little, too late to have any real effect now. Big Data is in full swing, and these latest deals are just companies realizing they have more to gain by charging other corps for direct access rather than letting them scrape or scrape-by-proxy.
Click to expand...
Soon it’ll all be gibberish either way considering some social media platforms are literally either filled with bots or people from India.
 
  • Like
  • Agree
Reactions: Ether Being, madethistocomment, lolcow yoghurt and 5 others
Miss Violet Beauregarde said:
Are ai users really that stingy with their prompts and settings that they're willing to call their friends to bum-rush the poor guy who just asked for the prompts?

Oh wait, i think i got it the wrong way
Click to expand...
Some AIfags are super stingy with their prompts. 4chan's AI generals on almost every board always have people complaining about their prompts getting 'stolen'.
 
  • Thunk-Provoking
  • Agree
  • Autistic
Reactions: Justwandering02, Ether Being, lolcow yoghurt and 5 others
Miss Violet Beauregarde said:
Got some Grade A seething over here, courtesy from our friend Tiktok

View attachment 5776617View attachment 5776618View attachment 5776619View attachment 5776620View attachment 5776621View attachment 5776623

btw is there any new hotness happening in AI? Only thing that came to mind was stable cascade and Mistral Large. Almost forgot that Sora video ai thing. Seems that the rate of cool new ai stuff is slowing down, dont you think?

EDIT: Ive got another cope coming up
View attachment 5776641View attachment 5776642View attachment 5776643View attachment 5776645View attachment 5776646
Click to expand...
1455744216633.jpg

Oh my GOD why would you ever fucking use red letters with purple outlining? It's fucking impossible to read.
This is getting MATI but the person who made that image deserves to lose their job to AI.
 
  • Agree
  • Feels
Reactions: phattie, Drain Todger, Ether Being and 9 others
DDBCAE CBAADCBE said:
I think Gemini AI is actually pretty decent. It’s one of the few I’ve found that can differentiate between different dinosaurs. The art isn’t always great but you get a gem here and there depending what you put in. It’s a fun little novelty if nothing else. Great for D&D character creation.
Click to expand...
What happens when you ask it to generate a nigersaurus?
 
Now for something different, an artist who's 1.) Inherently skilled, 2.) Embraces AI as an art tool, and 3.) Uses his own AI models using his old art as reference.


As I mentioned awhile ago, AI art will gradually eat jobs up the value chain, but a proper art director who can harness, manage, and direct it will always be valuable.
 
  • Informative
  • Like
  • Winner
Reactions: gorororo, Ether Being, lolcow yoghurt and 7 others
We will have to look out for mentions of his name on Xitter. He is bound to generate massive amounts of gooner and tranny cope.
 
  • Agree
Reactions: Kiwi2Shoes
  • Thunk-Provoking
  • Optimistic
Reactions: Ether Being, lolcow yoghurt and DavidS877
My updated list of how we should approach any content posted online, especially as it relates to news, "liveleaks", and other fringe media sources.

1. Verify if the event/moment/person has multiple different sources verifying the event/moment/person; preferably impartial posts (not family, friends, etc.).
2. Verify if said person announces public appearances or works at a physical job location
3. When investigating multiple sources, take in every fine detail such as mismatched clothing, objects not appearing, different lighting positions, perspective mismatch, and other environmental elements. AI face replication coupled with motion blur has become too good to be a reliable tell.

Decided to have this conversation with myself after watching this video:

 
  • Thunk-Provoking
  • Like
Reactions: Justwandering02, Ether Being, woowie queen and 1 other person
Absurdist Laughter said:
Take in every fine detail such as mismatched clothing, objects not appearing, different lighting positions, perspective mismatch, and other environmental elements.
Click to expand...

I like to go on reddit and accuse people of using AI in their art by making up inconsistencies that other people will confirmation bias into believing.

"That lighting on their left hand is *literally* impossible, can't believe you guys fell for this AI..."
"Umm... shadows don't work like that sweaty, nice AI!"

Makes for an entertaining 20 minutes tbh.
 
  • Like
  • Winner
  • Thunk-Provoking
Reactions: Secret Boots, Relinquish, N Space and 19 others
Absurdist Laughter said:
My updated list of how we should approach any content posted online, especially as it relates to news, "liveleaks", and other fringe media sources.

1. Verify if the event/moment/person has multiple different sources verifying the event/moment/person; preferably impartial posts (not family, friends, etc.).
2. Verify if said person announces public appearances or works at a physical job location
3. When investigating multiple sources, take in every fine detail such as mismatched clothing, objects not appearing, different lighting positions, perspective mismatch, and other environmental elements. AI face replication coupled with motion blur has become too good to be a reliable tell.

Decided to have this conversation with myself after watching this video:

Click to expand...
Right now most popular realistic AI checkpoints make near perfect bust-up images and selfies, with some generations where the eyes in general can fall into some uncanny valley, but you must be prepared and be expecting to see AI images...
Some more nitpicks are: hands in pants with only the thumb showing, kinda weird torso (if the image was a full body image, depending on the upscaler / background it might create an abnormally large torso) and generally boring compositions as perfectly in the middle of the frame, uncanny perfect background or boring poses like just standing, just looking ahead, etc, though that can be bypassed if someone puts in some effort and uses a controlnet to pose them.
There's also a inconsistency between AI generations of women and men (wow most people train models to generate only women for gooning, and also because there's more data available, who would've known?)

Additionally, the video you linked shows people checking real world printed photos, which of course doesn't let you zoom a 1000% in to check how the pixels are or if there is some camera defects, so they had it harder.
But generally you can get away with some jank shit if you do stuff like adding noise / filters and maybe you could entirely bypass faulty or inconsistent face gens in the case you were making a fake profile by just doing what some real people do and hide the face behind a emoji.
Normalfags and boomers are going to absolutely eat it up
 
  • Agree
  • Like
Reactions: lolcow yoghurt, woowie queen and Absurdist Laughter
You must log in or register to reply here.
Back