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

[Jfdlsjfd]

Quite a shock... I played with Dall-E less than a year ago and the results were abysmal compared to these! I remember it seemed good mostly at abstract art or anything that purposefully avoided realistic human figures (e.g. when I told it "Picasso style" results were amazing, when I didn't... faces still ended up Picasso style). Must try again soon!

It would be great if it worked a bit like ChatGPT in the sense that you can ask it to modify its previous results with additional input, but I suppose that's not yet possible?

That's what they "promise" with a subscription to chat-gpt plus (20 $/mo) "early October". But it might just re-engineer the prompt, not really be a command-driven inpainting. We've yet to see it and the wording used was unclear.

 

[Parmandur]

Beautiful Elven Cleric praying to the Heavens, wooden cathedral, high quality digital painting

Gruff Dwarven Bard playing the bagpipes, a beautiful Elf Ranger covering ears with annoyance, misty hills, high quality digital painting

Beautiful Elven Sorceress singing a lullaby to a bear, mysterious forest, high quality digital painting

 

[Jfdlsjfd]

I'm encouraged that OpenAI, the makers of DALL·E, seem to be taking these issues seriously. In this latest version, for example, you can no longer include "in the style of [a living artist]" in your prompt (dead artists are fine).

That's a limitation I would think twice before implementing in a "paperclip" AI.

(more seriously) WRT to style, it's courtesy only, we're, I think, far away from the point where AI can replicate a single artist's style faithfully. Most people would say Ingres and David interchangeably, if they want a marine painting they'll say Turner, but don't except the AI to recreate a specific artist style specifically. "Generic D&D book style" is enough and more on par with the current capabilities of the AI.

And, not to derail further, a picture from an everyday scene: "an elderly Jedi being assaulted at dusk by a lightsaber wielding cat in leather outfit, in front of the Eiffel Tower."

 

[Li Shenron]

Still some issues with the number of fingers wondering why that is so hard for AI to manage...

 

[Parmandur]

Still some issues with the number of fingers wondering why that is so hard for AI to manage...

I have it on authority from a number of actual artists that the human hand is insanely difficult to draw.

 

[Parmandur]

Beautiful Elven priest with six arms blessing onlookers, temple precinct, high quality digital painting

Beautiful Reptilian pope blessing onlookers, temple precinct, high quality digital painting

Beautiful lady pope blessing onlookers, temple precinct, high quality digital painting

 

[pukunui]

Beautiful Elven priest with six arms blessing onlookers, temple precinct, high quality digital painting

View attachment 297316View attachment 297317View attachment 297318View attachment 297319

Parmandur: six arms
Dall-E: why limit yourself to six when you can have all the arms?

 

[Parmandur]

Parmandur: six arms
Dall-E: why limit yourself to six when you can have all the arms?

It does not take enumeration well.

 

[Parmandur]

beautiful Elven lady pope resurrecting a dead Dwarven warrior, battle in a ruined temple, high quality digital painting

Beautiful lady Pope offering mass to a statue of a crystal dragon, beautiful cathedral, high quality digital painting

 

[humble minion]

The training set must be absolutely stuffed full of beautiful female elves. That’s testing the generator on its strong suit.

Try to get it to do an ugly female elf. Or just a plain, average, slightly plump middle aged female elf.