D&D General Would It Matter To You if D&D Books Were Illustrated by AI Instead of Humans?

[EzekielRaiden]

Perhaps a better way to phrase my concerns:

The phrasing of the question implies that regular, paid artists did not participate. That the system involved, at most, technicians from the tech company that runs the machine-learning art generator, and an art director from WotC, hashing out which pieces to use and then trimming them to the appropriate size.

That's the meaning I get from "illustrated by AI instead of by humans." Humans take no meaningful part in illustrating the document; AI completely replace all artistic effort other than layout.

 

[doctorbadwolf]

That's why I wrote "idea of being higher quality". I could have been clearer. It is not always a fact that handmade = better, as you justly pointed out (humans can make mistakes, or just don't care, or, in the case of very cheap handmade product, rely on reputation of handmade products without putting the effort). I think the analogy is pretty apt with artists in the future. For "run-of-mill" needs, if an AI can make you a black dragon picture to print and use as a token, or create a reasonable likeness of a character to put on a character sheet, the industrial product (easily available, low price) will be the most common opportunity. But a human artists, possibly using AI tool to make part of the work, will coexist with this offer, potentially offering a better quality (in the sense of "more fitting") product for when you need a very specific portrait for a campaign you know will last years or because you want your battlefield map to feature a very specific combination of things, or you really want a group of goblins to wear a specific uniform and so on.

Sure, that makes sense.

 

[Benjamin Olson]

So having gotten way into AI art the last week (my avatar is now an AI art assisted rendering of myself, ). Whole-cloth art generation with AI is a long way from being ready to sub in for art suitable for a published RPG outside of certain narrow use cases. However AI assisted reprocessing of images and compositing of AI generated variations of images allow someone with basic digital image manipulation competency to, starting from a rough sketch, photograph, piece of ai generated art, or a composite of all these things, create an image of a quality that historically would have required a proper artist, in a matter of 3-4 hours of trial, error, fine-tuning, and touching up (occasionally much faster, if it just goes your way and it's a simple subject and composition). Not only that, but in the process you get to try out many subtle and not so subtle variations, and sometimes encounter some inspiring improvisations from the AI. Here's some heavily AI assisted art I've made in the last few days using Stable Diffusion, each starting from a photo of a person in the rough pose I was going for:

Each of the people took 3-4 hours or running Stable Diffusion on an image, getting 6-24 variant outputs, compositing together the several best outputs, doing manual touch-ups, and running it through again. The Unicorn came together in about 45 minutes with minimal work from me; the program just nailed it. The gentleman at the top started as a photo of a football player, the lady in the center was a photo of a model in a miniskirt, and the unicorn was a photo of a horse.

Now there are things I dislike about each of these images, and each could probably use another pass. And the resolution is fairly limited due to my computer's VRAM. But if a guy who's qualifications are "kind of knowing his way around a photo editing application" and "having a decent, mid-range gaming rig" is producing stuff like this about a week after learning that AI art was even a thing, I'm confident that someone with actual talent or skill, and/or someone with more AI art generation practice, on a more powerful computer could probably fulfill most art needs of the RPG industry, and most other industries.

I don't think the future is AI stealing the jobs of artists. I think the future is AI assisted artists and not-quite-artists displacing proper artists actually producing things from scratch by being able to produce output faster and, sometimes in certain ways, better. This is the industrialization of art, and like it or not it will be ubiquitous sooner than you think.

 

[aia_2]

I got an update from Drivethru last week (iirc) where there was a new guideline for submissions with artworks AI generated. It is now necessary to disclose the presence of AI art at the submission but (...there is always a "but"!) in case the outcomes of the AI are significantly post-edited, then there is no need to such a disclosure (basically if the creator of the AI artwork changes it in a "significant" measure, he/she becomes the human author to the eyes of Drivethru, if i understood correctly)...

 

[Patrick Lewis1]

View attachment 260028

AI image generators such as Dall-E 2, Midjourney, and Stable Diffusion allow anyone to enter in a prompt for an AI trained on potentially billions of images to have it produce artworks that would take human artists far longer to create. Though they currently have shortcomings such as difficulty getting hands right, the technology is advancing quickly enough that even now there are D&D DMs and players using AI image generators to quickly create reasonably good images for their characters, NPCs, and locations.

The recent Spelljammer book was full of art, with WotC noting the higher than normal art budget. However, reviewers have bemoaned the lack of other content even as they appreciate the visuals. If they cared to, WotC could one day soon utilize an AI image generator to produce nice looking images with a much, much lower art budget. For example, they could hypothetically use an AI art generator to create images in the style of DiTerlizzi for the upcoming Planescape book.

Professional artists are obviously worried about being replaced, but for the end consumer will it matter who or what made the illustrations in a D&D book? I'm curious to see where opinions lie on this here.

Having Run Cyberpunk games alot.... I'd say this doesn't end well. AI will be writing games too soon. Just look at what YouTube's algorithm has done to our politics and our standards in public life. AI content is the removing of words from the dictionary of 1984....it is the lessening of our creative capacity.



But I want to be clear that I personally welcome our new robot overlords.

 

[Benjamin Olson]

I'll also add that each of the images I posted above has a couple of elements that don't quite measure up for reasons of being things AI particularly struggles with, that I am now hyper-aware of as someone who spent a lot of hours playing around with this stuff in the last week. In the near future we will all be better at recognizing the tells of even more carefully executed AI art, and non-AI assisted artists and their art will be particularly prized for their ability to handle the things that AI particularly struggles with, like proper hands properly holding things, complicated or irregular action poses, or non-human creatures that don't have clearly established common rules of how they are portrayed in art.

The particulars of what AI struggles with will doubtlessly evolve with the technology, but I think the basic principle that whatever AI can't master will become particularly prized and emphasized in traditional art as marks of authenticity holds sound. Thus even artists who never make any use of this tech and who, for whatever reason, find themselves secure from it as a direct economic threat will still find this technology profoundly affecting their art before too long.

 

[Art Waring]

Folks are thankfully having a reasonable discussion in this thread (thank heavens).

But as it stands, current legal precedent states that AI generated art is not protected by copyright law. You have to be a human being to create protected works. Anything you generate or use from an AI is public domain, and anyone else can use it freely.

"Under US copyright law, these images are technically not subject to copyright protection. Only "original works of authorship" are considered. "To qualify as a work of 'authorship' a work must be created by a human being," according to a US Copyright Office's report [PDF]."

 

[Stormonu]

Folks are thankfully having a reasonable discussion in this thread (thank heavens).

But as it stands, current legal precedent states that AI generated art is not protected by copyright law. You have to be a human being to create protected works. Anything you generate or use from an AI is public domain, and anyone else can use it freely.

"Under US copyright law, these images are technically not subject to copyright protection. Only "original works of authorship" are considered. "To qualify as a work of 'authorship' a work must be created by a human being," according to a US Copyright Office's report [PDF]."

That will change the first time a corporation uses it for their products.

 

[Art Waring]

That will change the first time a corporation uses it for their products.

Well, that's what we are waiting to see. ATM, legal precedent is lagging behind the times for AI law, but yes we may see the laws change in the future.

Currently though, at this time AI art is in the public domain. And so is the programming used to create these AI tools, so we will certainly be seeing more of them in the future.

 

[Sorcerers Apprentice]

Currently though, at this time AI art is in the public domain.

It remains to be seen if that is true if the matter is ever argued in court. I can see a lawyer arguing that the person providing the image generation prompts is actually creating the image, and the AI is just the tool being used.

I certainly wouldn't feel safe copying AI images other people have generated and using them in my own products, that's just asking for a lawsuit with an uncertain outcome.

What does seem reasonably certain is that the companies running the AI services do not have any copyright claim to the generated images, so if an AI generates an image for you, you should be good to use it however you want.