I’ll say it again because I always do. I’ve never seen AI art that I couldn’t tell was AI generated . It’s always wrong. The light source is always wrong, like the lighting is painted on subjects instead of cast onto them. And it lacks imperfections caused by human hand. Maybe it’s the photographer in me, but I’ve never seen believable lighting in AI art
You haven’t taken the time to train your own stable diffusion model on an artist’s work who is good at lighting. Shadow length and skew drawn by a suggested light source is pretty easy for SD models to start getting right, especially if they’re working from a gallery of one art style/type of composition. The article is stating what should be obvious to everyone at this point: this existed before the AI boom and you didn’t recognize it until the layman had access to the technology and didn’t refine the model or prompts to get these things right.
Oh I don’t work with the stuff, it’s just from what I’ve seen. I don’t really find any AI art I’ve seen to be this big impressive thing. I’m more interested in it from a data standpoint. I feel like not actually making your own art feels kind of depressing. Like what’s the point? Unless it’s for commercial use? Like if I feel creative, I’m going to make something.
Like I used to write, and I feel like if I wanted to write, I’d write. I don’t see the point in “writing” a prompt that pulls from other people’s work. Like what would I get out of it?
Yeah, commercial applications for it are great. It Makes life easier, lowers the barrier to entry, and hopefully will result in less work.
But for purely creative and cultural reasons, I just don’t see the point. Like I know nothing is original and we all pull from somewhere, but part of the enjoyment —to me— is the process of learning, of researching, reading other’s work to hone my craft.
And art without that is soulless and not an act of expression that comes from the deep reaches of ourselves.
It’s as empty as somebody buying a race car and a team to manage it, versus someone building their own and knowing every inch of it.
I’ll say it again because I always do. I’ve never seen AI art that I couldn’t tell was AI generated . It’s always wrong. The light source is always wrong, like the lighting is painted on subjects instead of cast onto them. And it lacks imperfections caused by human hand. Maybe it’s the photographer in me, but I’ve never seen believable lighting in AI art
Survivorship bias
You haven’t taken the time to train your own stable diffusion model on an artist’s work who is good at lighting. Shadow length and skew drawn by a suggested light source is pretty easy for SD models to start getting right, especially if they’re working from a gallery of one art style/type of composition. The article is stating what should be obvious to everyone at this point: this existed before the AI boom and you didn’t recognize it until the layman had access to the technology and didn’t refine the model or prompts to get these things right.
Oh I don’t work with the stuff, it’s just from what I’ve seen. I don’t really find any AI art I’ve seen to be this big impressive thing. I’m more interested in it from a data standpoint. I feel like not actually making your own art feels kind of depressing. Like what’s the point? Unless it’s for commercial use? Like if I feel creative, I’m going to make something.
Like I used to write, and I feel like if I wanted to write, I’d write. I don’t see the point in “writing” a prompt that pulls from other people’s work. Like what would I get out of it?
Yeah, commercial applications for it are great. It Makes life easier, lowers the barrier to entry, and hopefully will result in less work.
But for purely creative and cultural reasons, I just don’t see the point. Like I know nothing is original and we all pull from somewhere, but part of the enjoyment —to me— is the process of learning, of researching, reading other’s work to hone my craft.
And art without that is soulless and not an act of expression that comes from the deep reaches of ourselves.
It’s as empty as somebody buying a race car and a team to manage it, versus someone building their own and knowing every inch of it.
You’ve never knowingly seen AI art you couldn’t tell was AI art.