"Remove the people"
"What would you like them changed to?"
"... grass. I just don't want them there"
They can't do it.
isn't this supposed to be super easy to do with google magic eraser-type ai tools? hell i got decent at it using photoshop touch on an ipad before they killed it in 2015 (still mad)
Yah. The prompt 'engineers' don't seem to be skilled outside anything but prompt engineering, and it seems weird that even basic tools of the trade, so to speak -- let alone Googling how to accomplish it -- seem to be tough for them to manage.
yes it is, feels kinda fake to me
This is sort of the problem: nobody with even a modicum of skill in any form of art or asset manipulation wants anything to do with AI because it's objectively worse at performing the tasks they're trying to do, while also very explicit in its intent to make them homeless.
AI is literally only good for Step 0.1 in the PreVis pipeline, and anything after that, just use real tools.
Software engineers don't like to use AI not because of the intent to make them redundant but because it's s much more work to debug the resulting code.
It's absurdly funny that a company literally tried to hire "uh yeah I just babble tags at the inscrutable machine and it makes neat looking little things don't it" guys thinking that that's an actual skill set. They didn't even know how to use the tools that local AI generators have, they just gave it prompts lmao.
Only one of them even knew how to use Photoshop! And wasn't even any good at it! 🫠
I think this is the biggest aspect. If you're going to be an ai artist then digital photo manipulation is still a skill that is completely necessary.
That's not art, it's image manipulation. Anyone can learn to do it and it doesn't require creative talent, you just need to know the steps to follow to achieve x y or z in photoshop with a pre-existing image.
The issue is that even the people with that skillset don't really like AI that much. Content-aware fill being about as much AI as they wanna use. They can create nice stuff with collage snip and fitting multiple images together so they don't really need much AI.
You'd think "guy who parses what clients want into prompts the AI can create" would be one of the first jobs the AI would take anyway.
The whole value proposition of DALL-E 3 is that users can specify what they want in natural language, without prompt engineering, so this is a job that AI has already taken.
Idk how the guy with photoshop knowledge hasn't learned about inpainting, clone stamp, and content aware fill. feels fake off of that bit.
Imo the biggest benefit of AI is artists using it to speed up their workflow. I for example use it to mimic brushstrokes found in real oil paint art that would be insanely difficult to replicate in photoshop, fill in backgrounds, or to make a sketch I made while I was bored more than a sketch I forget about in a week
There are lots of people that do this prompting stuff with zero knowledge of computers or art and its very telling. If you train your own loras, use controlnet, inpaint with various tools, these things are all easy to fix. For me though, I hate taking orders when it comes to art, even when I'm doing it by hand, so obviously it will never be a job for me, just a hobby and sometimes people buy my shit.
Idk how the guy with photoshop knowledge hasn’t learned about inpainting, clone stamp, and content aware fill. feels fake off of that bit.
I'm a professional artist and I still struggle with a ton of that stuff. I don't really use photoshop much in my process though.
I like your idea of using AI to speed up workflow though. Could make things like texturing much quicker and easier.
I used to mostly do 100% pencil art when I was younger, but have since moved on to a weird hybrid of just taking quick pictures of my sketches and pulling it into photoshop to make it fancy / in color / whatever. So I've gotten really used to photoshop as a tool, so I guess if you're used to hands on stuff it makes sense that it would all be forlorn to you.
And yeah its very good at textures and so on. I love using it for skin texture, speeds the process up by an insane amount. If you are adding the skin on rather than generating a whole character, you can maintain cohesion for the character in between generations just with a bit of sketching. I don't think anyone has been able to spot the fact that my more in depth stuff had a ton of ai generations collaged together and painted over for this reason. When I'm being lazy and bored of course its a lot easier to spot, its basically just a fancier sketch
Seeing this and thinking about how I've been trying to get in position to make some money from art more frequently. As tools improve in ways that are more specialized than whole-image generation, I think incorporating them is starting to make sense for me. It's still a bit depressing, and I've been averse to the whole thing, but people who have an understanding of composition and the process already do seem to be situated to take advantage of this over general enthusiasts of the technology.
