D&D General DALL·E 3 does amazing D&D art

[Yaarel]

I've read the most recent 5-10 pages here but there are over 600 so I'm not sure I dare ask, as I assume this has been done to death, but I too am wow'd by a lot of this AI art, which seems so much better in the D&D domain than in other applications where I've seen it utilized, that said, the big question to me is what does this mean for fantasy illustrators and artists?

I think you are right that AI is better at illustrating familiar fantasy tropes, than other kinds of tropes.

Maybe fantasy requires artists to depict it, precisely because it doesnt happen in reallife.

So the AI is incorporating an interest and an abundance of fantasy artwork from reallife artists.


The corollary is, if a person is trying to innovate a new fantasy concept (or reconstruct a historical one) that isnt already part of popculture, the AI is less good at it.

 

[Thomas Shey]

I think you are right that AI is better at illustrating familiar fantasy tropes, than other kinds of tropes.

Maybe fantasy requires artists to depict it, precisely because it doesnt happen in reallife.

So the AI is incorporating an interest and an abundance of fantasy artwork from reallife artists.


The corollary is, if a person is trying to innovate a new fantasy concept (or reconstruct a historical one) that isnt already part of popculture, the AI is less good at it.

Some of it may be that even before people started playing with AI, there's been a pretty large amount of fantasy artwork out there (probably a combination of book, TV and movie fanart and stuff generated by D&D-adjacency in the first place). This is very noticeable when looking for art assets for NPC and monster (and even PC) tokens; its much easier to find fantasy than SF art, and easier to find SF than modern (non-urban-fantasy) art. As such there's a lot more teaching material to be had.

 

[Yaarel]

Of the two Wyvern images, the MM one is more animation and the PH one is more surrealism.

Personally, I always prefer surrealism styles, because I want to get a sense of what it would actually look like if it would be actually happening.

But there is room for "evocative" expressionism, anime genre, black-and-white humorous cartoons and pen-and-ink atmospheric sketches.

Especially core D&D does well to take an eclectic approach. But a specific setting can specialize in a signature look.

 

[Yaarel]

I am trying to understand an expressed difficulty with this Tiefling Bard.

The artwork is technically high quality. Good composition, vibrant use of color, dramatic use of lines and perspective, and so on. For my tastes, it is realistic enough for me to appreciate.

Obviously the portrait itself is appropriate for D&D fantasy, a Tiefling Bard with a medieval troubadour concept.


Is the problem that the clothing can come across as androgynous?

This kind of "dandy" or "metrosexual" or whatever one wants to call this kind of transgenderism exists in every culture, in every era, since the origins of the human species. There are many ancient examples of this.

This androgyny often happens in dreamlike contexts, especially entertainment and descriptions of the sacred.

(We know from modern examples of celebrities who enjoy these kinds of looks, that they are often successful with women. In every era there seem to be enough women who appreciate this fashion.)

There is every good reason for the D&D fantasy genres to include portraits of these humans who are part of the diversity of humans who populate our fantasy worlds.

 

[Nightfly]

what does this mean for fantasy illustrators and artists?

Nothing good, I'm sad to say.

 

[Lanefan]

Also, in our last session, we needed some diversion while a comrade was leaving the city unnoticed. In a ruthless artificer's way, we concluded that a character should select a rock, cast Reduce on it so it can fit into a bag of holding, take to the sky, empty the bag of holding, stop concentrating on Reduce by casting Enlarge. Since the maximum weight of a bag of holding is 250 kg, that would mean dropping a 4 tons rock from the sky. As illustated below.


A nice result for 2 2nd-level spells and a third level spell.

We do the same trick, only when the Reduced item is un-reduced it's usually way too heavy to hold or carry; you have to let go of it, meaning it has long since fallen out of range by the time you can resolve an Enlarge. (spells take measurable time to cast in our game)

Still a great way to sink ships, though - drop a large anvil from 500 feet and the decks/hulls of most ships are gonna have a hard time stopping it.

 

[Kichwas]

what does this mean for fantasy illustrators and artists?

Back in 1999 I got a job working at a Graphic Design company in the height of the first dot-com bubble.

Every one of the designers had been trained on print media, magazines, fashion shows, etc. They'd all taken a few classed in things like using an old 'pre OSX' Mac and using old Photoshop and Illustrator.

They had no clue how to do web page layouts. No one did. They had no idea about HTML and what you could or could not do.

I did, thus my job.

It was a new medium, new tools, and a new kind of client, and they had deadlines.

So, we set about learning stuff. I taught them the limits of the technology and as tools like Flash, JavaScript, CSS, and so on came online I'd run them through a 'you can now do this, but not that' set of sessions.

We made some ugly mistakes. sent out some bad files, and so on. But over time we mastered the new medium. That design house is still around today.

Back in the 1850s somebody figured out exposing light to celluoid or whatever and photography showed up. Now you could get a near instant portrait of the monarch of the day and - that was it for painters I guess. Except it wasn't. Folks took in the new tools, mastered them, and many artists will even blend photos and paint.

Some heretics even blend drawing and photoshop.

There's a new medium right now, and the old skills are not relevant to it. Folks will learn it, incorporate it some of the time and not at other times. Art will move on.

There's a LOT that can be done by mixing hand made work with AI assisted touch ups and AI added elements, back and forth, as just another piece of the process.

You can also just be the 'polaroid snapshot' equivalent and type in 'give me my stuff' and get some 'basic pics'.

People will learn to split the difference between art and 'quickie pieces' and things will move on.

 

[Yaarel]

People will learn to split the difference between art and 'quickie pieces' and things will move on.

I assume people will still appreciate human art. But I wonder what will happen when AI is genuinely artistic.

 

[Thomas Shey]

I assume people will still appreciate human art. But I wonder what will happen when AI is genuinely artistic.

It probably depends on what you mean by "genuinely".

 

[Kannik]

Here's a few of the things that bounce around in my mind when I muse on the impacts of these new generative models:

a) Though movies didn't fully take out theatre, and TV didn't fully take out movies, or how photography didn't fully take out painting, and etc, neither will generative models take out all work by artists/illustrators/etc. However, just like those precedents, they will make the market for those works smaller. Perhaps much smaller and more extreme than those precedents as both:

b) The speed of generation of these new techs is much faster than the precedents. Even photography... you still need to wait for or set up the best light, hike to the right location, pose the people, whatever. Sure, that might be faster than dragging out an easel and painting something, but compared to being able to bang out 12 iterations in 60 seconds it is still a long time. Sure, you still might need to go in there and inpaint and all that, but on the whole I think it's still much quicker, plus many companies might not care for that level of quality. We see bad photoshops today in professional situations, bad AI Art for pennies will be even more enticing.

c) The speed of adoption of these new techs is astronomical compared to the precedents. Whether we're talking iceboxes to fridges, or TVs from movies, it took time for the infrastructure to be built or for people to be able to purchase one / get one. Within the past 12 months anyone right now has access to a handful of different generative models. It's almost instantaneous. And, again, bad AI Art for pennies will be very enticing for many companies in cultures where money is the de facto "state religion". Thousands of illustrators and artists and photographers will find their "bread and butter" work eliminated almost overnight. Even if they get some great commissions from a patron now and again, that isn't living wage amounts of work. This could be economically devastating for many people. (While also impacting many other aspects including climate policies, infrastructure inequity, and more.)

d) Another big difference between these generative models and the precedents is that this work is not "creative." Movies were a new medium, but it still had creative humans behind them. Same with TV. Or photography. But generative models just put out stuff based on patterns it has seen in existing work. Sure, it will juxtapose some things some times that sparks something in our minds that we find cool and unexpected. But that's random happenstance, and the model itself doesn't recognize it. It may not spit that out again. And instead it will continue to generate variations on that which has been before (ie that it was trained on).

e) To which, thanks especially to the internet but it has always existed to some degree, there is a lot of aesthetic concentration that happens. Fantasy art that looks a certain way, expressions of form and bodies and etc that look a certain way, types of stories and the people who are in them that happen a certain way, and etc. Throw in a commercial aspect (as above with the de facto bit) and we get more and more of that as people try to follow the trend to ensure their product sells as much as it can. But it's a two way street -- if we are surrounded with the same type of imagery/work, we only grow to like that kind of thing (because it's all we've seen, or it's what we were first introduced to) plus we think that's 'normal' and even become socially concerned that others will mock us if we don't like it so we all get locked into that one aesthetic. Now throw in there that anyone and their dog can generate 100s of works in shorter time, and are also chasing the same aesthetic, and the models themselves are trained in that aesthetic and we start to get into a spiral/loop of limiting creativity and new expressions while also we as the viewers/readers get numb to the work. (It's a bit like doomscrolling -- we are seeing a lot but we are not getting anything out of it, we're making no connections.)



(Contextual disclosure: I am an Architect designing buildings for a living and therefore work in a creative field, and due to the communities I am involved in I have many artist friends.)