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


Used Bing to try and create a modern version of Mammon from his depiction in the 1e MM 2.
“Mammon is a red-gold color, and his scaled wings gleam like rubies. His form is bloated and soft looking, but Mammon is very strong. He is otherwise typically diabolic in appearance.”

Lastly my personal favourite
View attachment 296856

Question for folks using this as I dont have a Microsoft account and I'm quite shocked to see the variation in style.

Can someone ask it to do something in Shawn Woods style?

log in or register to remove this ad


CR 1/8
I haven't played with this stuff in quite a while, so it's astonishing to see how far this AI gen imagery has come in just the last year. A year from now, effectively all the little tell-tale oddities* in the outputs will be corrected, I bet.

* E.g., in MT Black's 1e PHB cover up top, the foreground greataxe (greateadze?) fighter appears to be waving around a... umm... a C-clamp maybe? I imagine him saying, "Hey Olive, you dropped one your thief tools!"

Soft, Watercolour?
You can play around with different combinations, and perhaps you will land somewhere in the right neighborhood! Bing Image Creator gives 4 result images every go, which tend to automatically be in different styles... but unfortunately, you cannot tell it 'more like this' when it does land on something that is almost correct.

Its use is more clicking to generate in the background while you're browing in other tabs.

Lucy is casting a Summon
Water elemental, and the teacher asked a Fire elemental.


A feral wizard is casting detect magic to identify if there is anything interesting in this ruin.


Opening a portal from Candlekeep to modern-day London.


Casting Smite Evil against a Sinful Woman.


An elven ruin, slowly being rebuilt by an entrepreneurial guild.


Checking for trap before: random statues are worrying in a tomb.

While Dall-e 3 is great at composition, it can be even better with a little work on it with regular tools. Too bad I don't have a powerful enough computer.


  • _e486bc5e-2745-4fdf-a40d-44121e28b2f0.jpg
    184.2 KB · Views: 688


On thing to note is that many of the improvements in the software doing what you were hoping comes from the prompt engineering side. Those people are tasked with converting user prompt language into things the systems actually understand. It is also how they keep a lid on NSFW elements and style theft. It's a really interesting job that literally did not exist a couple of years ago.

Improvement with dall-e 3 aren't that much with the image generation model (it's good, but it's not lightyears ahead, except on replicating text where it's really good) but with the ability to understand the prompt through analysis of a "natural language" prompt, thanks to the Chat-GPT 4 connection.

An Advertisement