I know classically we've used these types of illustrations to setup D&D adventures because it was easier/cheaper to produce, but now with Stable Diffusion you could try and get realism using control net from renders of images from this model for a twist on the classical. That could be cool.

Can this model generate maps for Red Alert 2?

Does anyone have a good inpainting model that works well with this?

This is a cool idea, but the CLIP guidance seems rather poor. Across a wide range of CFG values, it tends to produce mostly the same image. Even simple prompts like "castle" "cave" give most just the same run down hut in a meadow look. (v3)

Really love what I'm making with this! What art did you use to train the model? I'm wondering where I can find art to train my own dnd map model too

Very cool and awesome job thinking outside the box a bit on how to use SD.

Hello! this is my first time using this and I am a bit confused. I downloaded it, the 2GB.. but on what can I run it? How can I actually create? :D

When I use it, I often generate large empty maps, and the best results are only when drawing wilderness and ruins environments. When I want to draw a battle map that takes place in a town, no matter what hints I use about the floor and ground, the ground it generates is always yellow and green, just like in the wilderness. May I ask if there are any good prompt words that can make the generated buildings more dense and change the ground material? (This paragraph was generated by translation software. I apologize for any offensive or semantic errors.)

All prompts create a collection of near-identical outputs. Outputs are only wood ruins in a woodland setting in the shape of a circle within a circle.

Such is the outcome for the prompts:

stone, stone, stone, stone, stone, stone, stone, stone, stone, stone, 2d dnd battlemap

silver dragon, 2d dnd battlemap

field, yellow flowers, 2d dnd battlemap

road, 2d dnd battlemap

great in img2img. give it an image of something close and repeat as necessary.