Originally published at: Dungeons & Dragons puts the kibosh on AI art for its products | Boing Boing
…
This is such BS. And maybe he would have gotten away with it if he hadn’t turned in this obvious train wreck.
In general, I do not prefer the WotC D&D art style. It is (almost?) all digital and it just looks sketchy and unfinished most of the time. While ok for interior art, compare the 5E PHB or DMG covers to book covers by Easley or Elmore and tell me those are in the same league. They lack definition and detail.
But - putting that aside (art is subjective, and overall the illustrations are fine) - artists - especially established artists who have worked for the company before - turning AI art in is shooting themselves in the foot. If you turn in AI art for a gig, and everyone just goes along with it, next time they will cut out the middle man. You may think you have some secret process to use just the right special prompts to get the output you want - but you don’t. Not something that many other people couldn’t figure out. People already on the pay roll.
Maybe maybe it can be used to springboard something you work on top of - but unless the AI is learning from YOUR art, then you are ripping off other artists.
I saw people claim using various Photoshop tools is “AI” so what is the difference. Bullshit. Sit down with Photoshop and use those tools to create something. You can’t can you? It looks like shit. You can’t Heal Brush yourself into talent/skill.
The “tools” in Photoshop are just that. Yes, some of them do feel a little bit like cheating, but they aren’t inventing images whole cloth.
And note, I say Photoshop, but I know there are other and better programs for drawing digital illustrations.
(That said, I know Adobe is introducing actual AI systems in the newer versions, that is another bag of worms and I don’t think I like that either.)
Easley and Elmore yes, but also David A. Trampier and Erol Otus!
Tramp’s highly stylized black-and-white illustrations were amazing (as well as the full color Wormy comic which… ugh, how do we not have a hard cover omnibus already? Oh right… he died and his estate is probably a mess). And Erol Otus was just mind blowingly weird. Also black-and-white illustrations, also highly stylized, but so incredibly different and unique.
The fantasy genre definitely needs more artists working in different styles. Show the kids (and teens and adults) something new and mind-blowing!
WOTC keeps issuing proclamations like this after getting caught. It would mean a lot more if they actually managed to come out ahead of something like this instead of waiting until fans are pissed.
Course, at this point, I’m disinclined to give them any more money at all. I’m waiting to get a good look at Tales of the Valiant (that name sucks, Kobold) and maybe a couple other 5e clones/refittings.
The way I see it, there’s tools that help you get it done and don’t detract from your personal style (good), and tools that do it for you in their own style (bad). And tools like AI do it in a mishmash of many folks’ style (even worse).
I’m fine giving them the benefit of the doubt on this one. Catching AI generated art without crowdsourcing is a nontrivial hard problem
It’s an odd continuum to determine how much automation is sufficient. Different scenarios: using Gen AI to build the initial idea and then edit heavily with your own talents; drawing it yourself and then using AI to fill in texture details or to flesh out concepts in individual spots; or even taking an actual photo and using the tools to automatically remove components that you want gone and balancing the light to cover the difference (oh, wait, we already do this).
The blend between artists using tools and using their talent will prove hard to police. The real task is making sure that it’s the artists controlling their use of the tools and not publishers using the tools to control artists.
In this case, it’s not as bad as it might initially seem - the artist used an “AI” tool to alter an existing set of images he created. The AI wasn’t remotely generating the images wholecloth, it was adding details to some parts of the image (e.g. the original hands were quite sketchy, and the AI cleaned them up).
Of course, there’s still the issue that the tool works due to wholesale theft of artists’ work, and the resulting images are pretty generic - but his initial images and style were too.
Do you have a source on what was and wasn’t AI generated, I haven’t seen that yet.
In the 4 or 5 images I saw, there were still several elements that looked off or bad in places. Some off them, like a foot, I assumed was at the edge of their art board and wasn’t intended to be seen, just cropped.
But like in my example, the weapon is just a muddled mess with nonsensical
details.
An Art Director should know enough to know what to look for to at least start asking questions.
Even if these were 100% drawn, there are technical issues I’d ask for changes on.
David C. Sutherland III or go home. True, never has an artist been more easily translated into stiff, awkwardly posed lead miniatures, but I love it. Elements of so many of his original monster designs still inform what is in D&D today. I love the earnestness of his stuff, It makes me just feel good to look at it.
I mean, yeh. But it’s a new field with a new skill set: AI prompt engineering and detection.
I think it’s really easy to assign fault while forgetting just how good these models are getting at fooling even expert humans. In a few years it’s going to be even harder to detect.
I would stop giving them any money, if it wasn’t for the fact that I haven’t given them any since early 4th Ed.
They lost my custom a long time ago and I have seen nothing that has enticed me back.
Alma Mater remains one of the most disturbing and interesting RPGS I have had the pleasure of reading (never actually played it…we were in high school, so it seemed pointless). Erol’s art was incredible and creepy.
There are a host of other reasons to feel disinclined to give them money. I’ll assume that they didn’t catch the AI art because they didn’t know to look out for it, but they keep having other issues that require pressure for them to address. Instead of funding a business like that, point your dollars at other games that are either very similar or very, very different.
… nostalgia is a hell of a drug
This all started with the artist posting his drawings along side what the “AI” did to them, and a lot of it didn’t seem that changed. I didn’t look at them closely, just noticed the hands in particular were quite sketchy and the “AI” refined them and filled in the details (probably the feet, too, and some texture on the weapon and armor). The general impression I got was that with more time, the artist would have made them look more or less like what the AI did. The posts got deleted, but the images are probably still floating around.
The “tools” in Photoshop are just that. Yes, some of them do feel a little bit like cheating, but they aren’t inventing images whole cloth.
Current Photoshop maybe, but they are integrating generative AI tools into future versions right now. Supposedly the IP issues will be clearer as they are planning to train it only on the clip art that they own, but soon it will be able to generate pictures from a text prompt just like Stable Diffusion, DALL-E, etc.
I think the workflow that they are proposing is that generative AI will be used to generate the first draft and then the user will touch it up with other tools provided by the program.
Many of those original monster designs were from little plastic “dinosaurs” made in Taiwan that were used as miniatures. Things like the rust monster and owl bear.
I am curious to see that, too bad it got deleted. Interesting that AI drew hands better than the artists, hah.
Yeah, I pointed that out.
I’ve seen adds for it and Illustrator where it will come up with alternative color pallets for your design, for example. That alone isn’t bad, but the future integration with AI generated images, not a fan of.