[I am 99% sure] this is from a Shadiversity video (I went and looked for it, and I think he's unlisted it, because it's not in his list of videos anymore - the original 'castle'/AI video is gone, and only the swords video and second castle video - the collaboration with his brother - remain). I'm fairly certain his copy of Photoshop's primary purpose is to put a castle in the background of everything.
There was a segment in his first AI art video where he included some concept sketches for his comic book characters. This specific character also appears in the AI swords video briefly in a different pose - this time in an attempt to get the AI to put a sword in her hand.
It was an interesting video, because it actually showed the iterative process of getting Stable Diffusion to do something useful, touching things up, re-importing the image, setting it to only work on a specific portion of the image, and using low-freedom passes to blend foreground and background.
I tend to find the forced enthusiasm off-putting, but I was tinkering with Stable Diffusion at the time and the YouTube algorithm decided that Critical Drinker livestream + Stable Diffusion search = Shad that day.
I can still see it in his video list alright. For anyone who wants to check it out(it's a pretty good video on the topic): https://youtu.be/7PszF9Upan8
My hazy memories of the one from a month ago showed more of the Stable Diffusion UI.
It did, you're not wrong. This is a new one from 2 days ago. That one was on swords. People were probably sperging in his comments on the other videos.
AI version has a better face, better hair, better muscle contours on the arms and knees, better wrist and waist orientations, and might just have decided a generic background was more suitable than putting a sepia filter over a real life castle to claim as oc.
He did. There's multiple takes. He ran it through Stable diffusion a lot. He did one on swords as well, then put an image of his own sword from his book project in there to see what it could come up with.
The AI version is excellent and the detail is so much better, complete with an original background that looks like mountainous terrain from China's rural provinces.
The original was drawn by a sewer gollum that's never seen the real world or living people that it wasn't having for dinner.
Which is the problem they are having. "Artists" have had so little collective talent that many of them can't even reach "generic" level, and are now upset that the bar has been raised on them.
They also think things like "perfect anatomy" and "extra details" are so incredibly important that they fail to make anything worth looking at in pursuit of it. Wherein the AI may fail these tests, but are still appealing to look at.
The entire point of AI driven image generation is that you give it a bunch of images that look like a cat, you say "this is what a cat looks like", and it learn the elements that make a cat. It knows a cat has pointed ears, paws, a long tail, and so on.
The person that drew that is Shadiversity on Youtube.
He's never claimed to be an artist, that he used to draw a lot but was never very good at it. He's made multiple videos, including multiple lengthy livestreams, talking about AI art, what he's been able to achieve with it, as well as the limitations. He's said it's a useful tool for people to get into art, create it, or to elevate it without having to spend tons of time mastering it, or beyond their natural talent (or lack of it). He's had actual artists on his videos, both agreeing and disagreeing with the use of AI art.
He's also said with the current tech it still takes quite a lot of know how to use the software properly, hours to perfect it, and to get specific results, often requiring splicing parts together from multiple different renderings. Shad says he's using AI art to create characters for a new IP he's working on.
In one of the livestreams they accurately paralleled AI art to photoshop or sound synthesizers, enabling more people to get into photo manipulation and song orchestration, who wouldn't otherwise have the money and/or time to do so.
If you have the time, I highly recommend Shad's videos on the subject. He was introduced to it by his friend Oz.
Meh, I think the right is a workable improvement and I have seen worse pass for illustration. Granted, poorly utilized perspective will always look bad in ways that AI can't really fix, the legs on both of these hurt me. I cannot argue that the AI didn't improve this man's work - but that is no excuse to not get better, and this example makes it painfully clear.
Also, being conservative with detail is very much a design 101 thing and I am surprised that such a lofty name as RJ would even think to present the reduction of redundant detail as an argument against AI, but then again his work tends to be gritty reboots of pokemon ala Dave Rapoza - decent, but a dumb argument nevertheless.
Because it's a sepia filter placed over Neuschwanstein, an actual location in the real world. This is worse than tracing over a castle and then claiming it's oc.
The original is by a communist who recently claimed AI image generation software is meant to generate child porn in a deplatforming campaign, so all criticism he comes up with is a hysterical lie no matter what. The other person is at the top of the bell curve and repeating words they vaguely associate with "bad art on computer" as a concept.
Since you're also clearly at the top of the bell curve, you should try thinking really, really hard about what the word "original" means in regards to this screencap until it clicks.
The "original" artist just yoinked some photo of Neuschwanstein for the background. It's not like that was all his original work to begin with.
But it's okay when humans do it!
[I am 99% sure] this is from a Shadiversity video (I went and looked for it, and I think he's unlisted it, because it's not in his list of videos anymore - the original 'castle'/AI video is gone, and only the swords video and second castle video - the collaboration with his brother - remain). I'm fairly certain his copy of Photoshop's primary purpose is to put a castle in the background of everything.
There was a segment in his first AI art video where he included some concept sketches for his comic book characters. This specific character also appears in the AI swords video briefly in a different pose - this time in an attempt to get the AI to put a sword in her hand.
It was an interesting video, because it actually showed the iterative process of getting Stable Diffusion to do something useful, touching things up, re-importing the image, setting it to only work on a specific portion of the image, and using low-freedom passes to blend foreground and background.
I tend to find the forced enthusiasm off-putting, but I was tinkering with Stable Diffusion at the time and the YouTube algorithm decided that Critical Drinker livestream + Stable Diffusion search = Shad that day.
I can still see it in his video list alright. For anyone who wants to check it out(it's a pretty good video on the topic): https://youtu.be/7PszF9Upan8
Maybe this is a reupload? My hazy memories of the one from a month ago showed more of the Stable Diffusion UI.
It did, you're not wrong. This is a new one from 2 days ago. That one was on swords. People were probably sperging in his comments on the other videos.
Hmm. I guess my quick scan of the sword video for a timestamp missed what I was looking for.
AI version has a better face, better hair, better muscle contours on the arms and knees, better wrist and waist orientations, and might just have decided a generic background was more suitable than putting a sepia filter over a real life castle to claim as oc.
He did. There's multiple takes. He ran it through Stable diffusion a lot. He did one on swords as well, then put an image of his own sword from his book project in there to see what it could come up with.
The AI version is excellent and the detail is so much better, complete with an original background that looks like mountainous terrain from China's rural provinces.
The original was drawn by a sewer gollum that's never seen the real world or living people that it wasn't having for dinner.
That's how it works. Neural networks can only make a generic mashup of its training data. It's always going to look soulless.
Its proper use lies in 1. the equivalent of stock photography and 2. quick variations or initial templates for artists to work on.
Fuckin' amateur.
Ha! It's still generic though.
Which is the problem they are having. "Artists" have had so little collective talent that many of them can't even reach "generic" level, and are now upset that the bar has been raised on them.
They also think things like "perfect anatomy" and "extra details" are so incredibly important that they fail to make anything worth looking at in pursuit of it. Wherein the AI may fail these tests, but are still appealing to look at.
I agree. It's mediocrity and lack of inspiration makes it a little more difficult to distinguish from NPC writing, to be honest.
And non-AI art isn't?
The entire point of AI driven image generation is that you give it a bunch of images that look like a cat, you say "this is what a cat looks like", and it learn the elements that make a cat. It knows a cat has pointed ears, paws, a long tail, and so on.
Imagine drawing this and then having the audacity to call yourself an artist.
The person that drew that is Shadiversity on Youtube.
He's never claimed to be an artist, that he used to draw a lot but was never very good at it. He's made multiple videos, including multiple lengthy livestreams, talking about AI art, what he's been able to achieve with it, as well as the limitations. He's said it's a useful tool for people to get into art, create it, or to elevate it without having to spend tons of time mastering it, or beyond their natural talent (or lack of it). He's had actual artists on his videos, both agreeing and disagreeing with the use of AI art.
He's also said with the current tech it still takes quite a lot of know how to use the software properly, hours to perfect it, and to get specific results, often requiring splicing parts together from multiple different renderings. Shad says he's using AI art to create characters for a new IP he's working on.
In one of the livestreams they accurately paralleled AI art to photoshop or sound synthesizers, enabling more people to get into photo manipulation and song orchestration, who wouldn't otherwise have the money and/or time to do so.
If you have the time, I highly recommend Shad's videos on the subject. He was introduced to it by his friend Oz.
Meh, I think the right is a workable improvement and I have seen worse pass for illustration. Granted, poorly utilized perspective will always look bad in ways that AI can't really fix, the legs on both of these hurt me. I cannot argue that the AI didn't improve this man's work - but that is no excuse to not get better, and this example makes it painfully clear.
Also, being conservative with detail is very much a design 101 thing and I am surprised that such a lofty name as RJ would even think to present the reduction of redundant detail as an argument against AI, but then again his work tends to be gritty reboots of pokemon ala Dave Rapoza - decent, but a dumb argument nevertheless.
The background is better on the original. The AI generated person is better. What's so hard to understand here people. It's pretty fucking obvious.
The original's background is a photo.
Because it's a sepia filter placed over Neuschwanstein, an actual location in the real world. This is worse than tracing over a castle and then claiming it's oc.
"original castle, donut steel"
Shadman yes
The original is by a communist who recently claimed AI image generation software is meant to generate child porn in a deplatforming campaign, so all criticism he comes up with is a hysterical lie no matter what. The other person is at the top of the bell curve and repeating words they vaguely associate with "bad art on computer" as a concept.
Since you're also clearly at the top of the bell curve, you should try thinking really, really hard about what the word "original" means in regards to this screencap until it clicks.
You're the one who started snarking first. Fuck off.