Bringing D&D/AD&D campaign settings to life with Stable Diffusion

[Zed Duke of Banville]

Cyberpunk artwork in the style of Syd Mead, Neil Blevins, Masamune Shirow, and Ron Cobb:

 

[gurugeorge]

Fuck's sake, the rate at which this is accelerating most of the digital artists/concept artists/so on will be begging on the streets in like, what, 2 to 3 years.

 

[Zed Duke of Banville]

I'll start worrying when AI learns how to button-rate properly.

 

[Blutwurstritter]

Cyberpunk artwork in the style of Syd Mead, Neil Blevins, Masamune Shirow, and Ron Cobb:

Looks great.

 

[Lagole Gon]

 

[Lagole Gon]

Hmmm...

 

[Mortmal]

Hmmm...

View attachment 43406

Thats more arcanum than D&D, but nice.

 

[Old One]

How many human women died horrible deaths to bring us that image?

 

[Zed Duke of Banville]

Homages to Jim Holloway and Stephen Fabian

Spoiler

Adventurers arguing over directions, from The Wilderness Survival Guide (1986)
Itheldown Castle, from GAZ11 The Republic of Darokin (1989)

 

[mediocrepoet]

Homages to Jim Holloway and Stephen Fabian

Spoiler

Adventurers arguing over directions, from The Wilderness Survival Guide (1986)
Itheldown Castle, from GAZ11 The Republic of Darokin (1989)

Spoiler

Woman yells at cat, from The Internet (2019)

 

[Mortmal]

 

[Alex]

Homages to Jim Holloway and Stephen Fabian

Spoiler

Adventurers arguing over directions, from The Wilderness Survival Guide (1986)
Itheldown Castle, from GAZ11 The Republic of Darokin (1989)

Spoiler

View attachment 43530

Woman yells at cat, from The Internet (2019)

My first thought looking at the image was that it was a mexican standoff with magic missiles instead of guns.

 

[gurugeorge]

I've just noticed the guy on the left is pointing with his middle finger, and the guy on the right in the background, God knows what's going on there - with both hands.

It's weird how AI still seems to be a bit baffled by hands. Eyes are getting better, but you still see a fair amoutn of googly eyes in AI art.

I wonder how much of the AI kerfuffle parallels the old kerfuffle around photography and the supposed redundancy of painting. If the AI isn't actually smart, then there's no real worry there, it's just another medium that artistic-minded people can use (after all they have to train and select, to get used to the medium's funny little ways, same as with photography).

So is the result going to be just more art, both mediocre and good quality? (Actually the result will probably be the devaluing of a certain kind of illustrative art.) I know there's a whole thread on this so I'll stop there.

 

[Darkwind]

it's just another medium that artistic-minded people can use (after all they have to train and select, to get used to the medium's funny little ways, same as with photography).

So is the result going to be just more art, both mediocre and good quality? (Actually the result will probably be the devaluing of a certain kind of illustrative art.) I know there's a whole thread on this so I'll stop there.

EDIT: Not to derail further, since there is another thread but, briefly...

Disagree completely. Art required a certain amount of skill and dedication to the craft. This was generally developed painstakingly over years of practice.

Now any midwit can simply dick around with prompts and once the AI gets good enough you are generating art on par with some of the greatest artists of all time. Yes, the hands and eyes are super creepy but the pace is quickening and it won't be long until the images are flawless. At that point, any scrub who can fumble his way through some prompts, trial & error, and sheer outright copypasta can create fairly compelling art.

For the fantasy genre, as an example, why hire the next Easley, Elmore, Brom, Caldwell, or Reynolds when I can simply reproduce work that is 95% their quality at a fraction of the cost & time?

If I were a freelance artist I'd be experiencing existential dread right about now. But that goes for many jobs. Coders, lawyers, anyone that uses MS-Word as their primary tool for making a living is in very serious danger in the next few years. You'd be foolish not to see this rapid progression at this point.

 

[Zed Duke of Banville]

K.u.K. Kosmonaut Zsigmond von Tichy, the first man on Mars

 

[Kev Inkline]

What checkpoints/loras are you using?

 

[Zed Duke of Banville]

What checkpoints/loras are you using?

These newer images, from mid-October onward, are created with Stable Diffusion XL. No loras or other extras, except for the built-in face restoration (not always used) and sometimes inpainting or image-to-image.

 

[mediocrepoet]

I've just noticed the guy on the left is pointing with his middle finger, and the guy on the right in the background, God knows what's going on there - with both hands.

It's weird how AI still seems to be a bit baffled by hands.

I didn't notice until you pointed it out, but I should have looked. AI hands are the stuff of nightmares.

 

[NecroLord]

Hmmm...

View attachment 43406

A tarantian gnome and a half-orc.
That doesn't look like a half-ogre.

 

[Joonas]

Somewhat related... I generated a big sketch art & lore book for my roguelike game Zorbus (thread here). The images are generated with Microsoft Bing (DALL-E 3).

Contents, some sample images, and a 140 megabyte PDF available here:
lore.zorbus.net

Online flipbook:
lore.zorbus.net/flipbook/

 

[Caim]

Lmao vagina cave.

 

[Bruuce]

Making art for an Edgy Cyberpunk World: Part 1: Random Dudes
I like having unique tokens for all my NPCs, to a level where it adversely affects my life.
Enter Art AI, no longer will I have to trudge through art sites and token packs full of clashing art, finally I can have proper art for everything!
I had a certain gritty comic book-like look in mind and ended up getting close enough,

Starting off by making random civilians on Midjounrey

The following was made with Midjourney v5, which I only got to play with for the day or so when they re-opened trials.

I got good result by using a vidya screenshot and "comic art by moebius" as part of the prompt.

Then midjouney killed all the free trials so, because the poorness that dissuades me from paying for the midjourney sub prevents me from getting a good grafix card, I went to try stable diffusion, finding a tolerable free site (which is also paid only now)


Comic Art by Moebius worked here too and gave me some good stuff.

Stable Diffusion Simps

Stable Diffusion Security Officers

Had to switch image hosts because imgur HATES women

Random Bystanders

I also made some wasteland mutants, using both the site and using my own computer but didn't get anything of worth.

More Stable Diffusion Artstyle Experiments

I did some more basic prompt experiments and it is pretty good at giving you a consistent style based on prompts like "art by Carlos Ezquerra" next to "cyberpunk mall cop" even if the result is only vaguely similar to the artists' actual style and only subtly different from the other artist prompts.

"cyberpunk mall cop, drawn by Simon Bisley"

"cyberpunk mall cop, drawn by Liam Sharp"

"cyberpunk mall cop, drawn by Brain Bolland"

"Cyberpunk mall cop, art by Pat Mills"

Prayin' that this loads properly.

 

[mediocrepoet]

Lmao vagina cave.

 

[Lagole Gon]

 

[Caim]

As it turns out Mr Clean's first name is Veritably, which makes him sound like a Discworld character.