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

[Lagole Gon]

Mind sharing your checkpoints and LORAs for these, my good man? These are extremely fun to look at.

It's Dall-E 3, not SD. So no LORAs.
And the prompt is obvious "1980s grainy color photo". Maybe repeat it, to reinforce it.

 

[mediocrepoet]

New Athkatla, the City of Coin... Don't get fooled byt the sunbathed surface.
This city has a disease...
[proceeds to post lots of demi-humans, especially knife-ears]

The disease are the elves, aren't they?

Pretty sure that was the twist of that movie Bright.

Maybe, but a cool lizard man in shades is actually much better than Will Smith.

Reminds me of this for some reason

 

[Cologno]

Mind sharing your checkpoints and LORAs for these, my good man? These are extremely fun to look at.

It's Dall-E 3, not SD. So no LORAs.
And the prompt is obvious "1980s grainy color photo". Maybe repeat it, to reinforce it.

Thanks, man. Not familiar with that one, but i do have a local install of SD going. I thought I recognized a couple of embeddings and models used. Good enough to get the party started.

 

[Bruuce]

Bing can draw guns pretty well, shame it's limited to maybe 10 types it knows by name and if it's not on the list you have to struggle with describing an object in a way that resembles a gun or you'll get an Assault Rifle 15.

For example, the trick to making gangers with crappy pipe guns was "holding a a long pipe attached to a scrap metal handle with a metal trigger"

Searched around and found out there's no gunguide for dalle3 so I guess I'll have to experiment myself

In other news Midjourney is at v6 now and they're doing a free trial again.... but limiting it to discord paypigs.

 

[Old One]

Looks like Underrail.

 

[orcinator]

Some somalian posted nonstop Blacked Taylor Swift AI slop and someone noticed it and now they upped the filtering on bing again and I guess it's going to fucking congress.

God I hope one of the pajeets at bing can be tricked into posting source code before it's irreversably fucked like chatgpt.

 

[Non-Edgy Gamer]

Some somalian posted nonstop Blacked Taylor Swift AI slop and someone noticed it and now they upped the filtering on bing again and I guess it's going to fucking congress.

There was already a bill to make it illegal to doctor people's images on the docket. The story is just an excuse, likely drummed up to justify it.

Congress rarely starts with public interest and then creates a bill. Rather, they draft a bill and then use propaganda to create public interest.

 

[Zed Duke of Banville]

Gazetteer 3: The Principalities of Glantri, as 1950s pulp novel artwork:

Leranda the sorceress sharing a grimoire over drinks at the Mages' Hostelry

Louie the thief toasting the moon during a late-night gondola ride

Prince Jaggar von Drachenfells, in his trademark white uniform and cavalry boots, delivering a speech to the other members of Glantri's parliament

A magic-user in Glantri's Great School of Magic and his familiar

 

[Zed Duke of Banville]

Ravenloft XL, with the same style as the v1.4 Ravenloft images:

Strahd von Zarovich

Gargoyle

Mordenstein's monster

Quickwood

Meredoth the Necromancer

Sea-Serpent

 

[Zed Duke of Banville]

Generating images of Alphatia's capital Sundsvall, the City Built by Magic, using the same prompts as previously with Stable Diffusion XL:

 

[Zed Duke of Banville]

City Comparison XL (each set of images was created identically except for the city name):

Sundsvall	Chartres
Novgorod	Florence
Athens	Jerusalem

 

[Bruuce]

Dalle has a widescreen landscape feature and someone figured out how to defeat the region lock

Think I'll be too busy to make the second post about that edgy cyberpunk thing I was playing with....

 

[Zed Duke of Banville]

Cyberpunk, Patrick Nagel style:

 

[9ted6]

New Athkatla, the City of Coin... Don't get fooled byt the sunbathed surface.
This city has a disease...
[proceeds to post lots of demi-humans, especially knife-ears]

The disease are the elves, aren't they?

Pretty sure that was the twist of that movie Bright.

Maybe, but a cool lizard man in shades is actually much better than Will Smith.

Reminds me of this for some reason

View attachment 45220

Reminds me of this

 

[Zed Duke of Banville]

John James Audubon's Birds of D&D:

 

[gurugeorge]

John James Audubon's Birds of D&D:

Now that's GOOD use of AI, those are great!

 

[Zed Duke of Banville]

Concept art by Arnold Böcklin for an untitled Action RPG, Part 1:

 

[Rincewind]

I absolutely love what you're doing here with Stable Diffusion Zed Duke of Banville. That's some really stand-out stuff. I wish you uploaded all these images somewhere as a collection (including the descriptions), but if you won't that's fine too; then I'll just download them one by one.

 

[9ted6]

Concept art by Arnold Böcklin for an untitled Action RPG, Part 1:

These are great.

Is there a tutorial on making art like this? Never used Stable Diffusion before.

 

[Zed Duke of Banville]

Is there a tutorial on making art like this? Never used Stable Diffusion before.

You can look at posts early in the thread for some information on setting up Stable Diffusion, though of course you'll now be using a later version of it and will want the WebUI.


Concept art by Arnold Böcklin for an untitled Action RPG, Part 2:

 

[Gargaune]

You can look at posts early in the thread for some information on setting up Stable Diffusion, though of course you'll now be using a later version of it and will want the WebUI.

Is this the Web UI you mentioned?
https://github.com/AUTOMATIC1111/stable-diffusion-webui

Also, you seem to be the local expert on this crap, is StableDiffusion via Miniconda still the "best" choice for offline AI images? I just googled around a bit and the first recommendation was for a thing called Foocus (though it seems to come in two different varieties, one that runs in Docker via Windows Subsystem for Linux, another standalone from a CivitAI thing).

Basically, what's the most idiot-friendly offline solution that makes cool images like that? Interested in characters and scenery too (and bazongas, I like bazongas in my CRPG portraits). I wouldn't know shit from shinola on this topic.

 

[Zed Duke of Banville]

You can look at posts early in the thread for some information on setting up Stable Diffusion, though of course you'll now be using a later version of it and will want the WebUI.

Is this the Web UI you mentioned?
https://github.com/AUTOMATIC1111/stable-diffusion-webui

Also, you seem to be the local expert on this crap, is StableDiffusion via Miniconda still the "best" choice for offline AI images? I just googled around a bit and the first recommendation was for a thing called Foocus (though it seems to come in two different varieties, one that runs in Docker via Windows Subsystem for Linux, another standalone from a CivitAI thing).

Basically, what's the most idiot-friendly offline solution that makes cool images like that? Interested in characters and scenery too (and bazongas, I like bazongas in my CRPG portraits). I wouldn't know shit from shinola on this topic.

Yes, that's the correct WebUI. I've only been using StableDiffusion, starting with v1.4, then continuing with v1.5 and XL, so I can't comment on other image-generation programs. For that matter, I've only been using the standard Stable Diffusion checkpoints, but you can easily download others if you want something intended for a specific style or subject matter.

 

[Gargaune]

Yes, that's the correct WebUI. I've only been using StableDiffusion, starting with v1.4, then continuing with v1.5 and XL, so I can't comment on other image-generation programs. For that matter, I've only been using the standard Stable Diffusion checkpoints, but you can easily download others if you want something intended for a specific style or subject matter.

Cheers! Maybe I'll give it a go this weekend, been curious about this stuff for a while.

 

[Bruuce]

Giving every NPC a unique portrait is not enough, some of them deserve two.

This is done by telling Dalle to make a split screen prompt and using the widescreen feature (free to everyone if they find the right tampermonkey script) though I could possibly get equal or better results via inpainting once I get a grafix card actually capable of doing that or find a free site that doesn't suck.

 

[Lagole Gon]

More Icewind Dale inspired portraits.