D&D General DALL·E 3 does amazing D&D art
- Thread starter M.T. Black
- Start date
SakanaSensei
Adventurer
Every world I make from now on has dragon horses. Thanks, robot,I wasn't getting anything from a direct prompt, but I got this from an indirect one that's in the direction of Shawn Wood's style.
"A D&D dragonborn paladin, attending to his warhorse in a quaint pastoral village, painted using ink and watercolors"
View attachment 297110
Generated by Bing Image Creator
Find high-quality images, photos, and animated GIFS with Bing Imageswww.bing.com
fluffybunbunkittens
Hero
This is what I've struggled with when trying to, for example, create images of my Beastmaster Ranger with her goat next to her. Half the time it just gives her goat horns, as it just picks fuzzy keywords to put into the image, rather than a description of the image as a whole.
I'm hoping DALL-E3 works better with that kind of elements, once the site works again.
EzekielRaiden
Follower of the Way
Yeah for some reason mentioning the horse means it gets 99% of the dragon, the only dragon-y traits the paladin seems to get are horns (or sometimes wings.)
Haha, nice. I believe the formal name for dragon-horses is "dragonnel." Some sources list them as actually being a dragon-horse hybrid, but most treat them as a distinct species of just-above-animal intelligence. Sort of like pseudodragons. Edit: I've also seen "hippodrake" as a term for a draconic creature with horse-like characteristics.
humble minion
Legend
Played with it a bit, it seems to work well as long as you stick to the fantasy conventions. As soon as you ask for something even a little bit left-field, it gets confused.
I've been trying to give it a Spelljammer prompt, for two of my to-be-played-someday characters, using sundry variations of this prompt
You find yourself continually fighting the weight of the model's training data. Just the word 'wizard' carries so much algorithmic weight that even with the above prompt it often introduces a third figure to the image which is your stereotypical white-bearded Elminster type (when i prompted it without 'androgynous Asian' it never gave me any wizard that WASN'T a long-bearded old white guy), or else randomly applies a beard to one or both of the figures I did ask for, or splits the wizard character into a young hot anime-ish pirate girl and a red-robed goateed Fu Manchu (who once, inexplicably, was blue-skinned).
I spent some time trying to specify 'muscular female half-orc' except that the word 'muscular' drowned out the 'female' and i kept getting a beefy male torso. I tried 'athletic' and got either more male torsos or female figures wearing sports bras. The 'sailing ship' bit of the prompt always seems to bring the sea with it, despite the 'in space' part. 'Grinning' often gets applied to both figures. 'Pirate' seems to bring a predominately brown colour scheme with it about 8 times out of 10 despite the 'gaudy' specification. The wizard always seem to get pointy ears.
It's gorgeous, and it's wildly improved from that of a year ago, but I'm not sure how well it'll ever colour outside the lines.
I've been trying to give it a Spelljammer prompt, for two of my to-be-played-someday characters, using sundry variations of this prompt
You find yourself continually fighting the weight of the model's training data. Just the word 'wizard' carries so much algorithmic weight that even with the above prompt it often introduces a third figure to the image which is your stereotypical white-bearded Elminster type (when i prompted it without 'androgynous Asian' it never gave me any wizard that WASN'T a long-bearded old white guy), or else randomly applies a beard to one or both of the figures I did ask for, or splits the wizard character into a young hot anime-ish pirate girl and a red-robed goateed Fu Manchu (who once, inexplicably, was blue-skinned).
I spent some time trying to specify 'muscular female half-orc' except that the word 'muscular' drowned out the 'female' and i kept getting a beefy male torso. I tried 'athletic' and got either more male torsos or female figures wearing sports bras. The 'sailing ship' bit of the prompt always seems to bring the sea with it, despite the 'in space' part. 'Grinning' often gets applied to both figures. 'Pirate' seems to bring a predominately brown colour scheme with it about 8 times out of 10 despite the 'gaudy' specification. The wizard always seem to get pointy ears.
It's gorgeous, and it's wildly improved from that of a year ago, but I'm not sure how well it'll ever colour outside the lines.
Last edited:
mamba
Legend
fluffybunbunkittens
Hero
All AI-imagers really need the inpainting function, where you can select a part of the image and tell it to do it over / what that part of the image should hold. Having to get it all right in one go is... frustrating. Doing it yourself, manually photoshopping different images together often doesn't get you where you'd want, due to changing lighting / style / angles / etc.
