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

[Princelucianus]

Seven friends playing D&D in a 17th century Dutch house. Style Rembrandt

 

[tsadkiel]

The AI figured out what I was up to.

 

[Parmandur]

The AI figured out what I was up to.

View attachment 333771

Alright, what was the prompt, lol

 

[tsadkiel]

Alright, what was the prompt, lol

classic Disney style, elven priestess of the moon with long silver hair, long yellow gown, dancing with hairy well-dressed beast, ballroom, chandelier, high angle

It was the Disney that got me, I think. I was trying various styles and they all looked weird.

 

[tsadkiel]

And the beasts weren't all so on the nose.

 

[tsadkiel]

Tale as old as time.

 

[Saracenus]

Fusion time... Samurai done in the portrait style of a medieval knight.

GENERATED PROMPT: An acrylic painting of a medieval Japanese samurai, reimagined in the context of a European knight's tapestry. The samurai, adorned in detailed armor, is depicted in a powerful, yet serene pose, holding a katana. The tapestry-style background weaves together elements of Japanese and European medieval culture, featuring motifs like cherry blossoms alongside European knights and castles. The texture of the tapestry is visible, with rich, deep colors and intricate patterns that highlight the fusion of these two distinct cultural art forms. The interplay of light and shadow in the painting accentuates the depth and dimension of the scene, fantasy artisan prompt, hd quality, natural look --ar 1:1

GENERATED PROMPT: A painting of a medieval Japanese samurai, depicted in full armor, in a classic European knight's portrait pose. The samurai stands in a dignified, stoic pose typical of knight portraits, with a katana sword held gracefully. The background features a blend of Japanese and European medieval elements, like a castle with cherry blossoms. The lighting is soft, reminiscent of classical European paintings, highlighting the intricate details of the samurai's armor and the serene landscape behind. Drawn with oil painting techniques, capturing the texture of canvas and brushstrokes, fantasy artisan prompt, hd quality, natural look --ar 1:1

 

[Greenstone.Walker]

How are you getting all this to look right? Every time I've asked DALL-E for art (we have an account through work) I get misaligned weapons (like bows with two or three arrows nocked), strange numbers of fingers, limb misalignment and other oddities.

 

[pukunui]

How are you getting all this to look right? Every time I've asked DALL-E for art (we have an account through work) I get misaligned weapons (like bows with two or three arrows nocked), strange numbers of fingers, limb misalignment and other oddities.

What version of DALL-E are you using? Most of us are using the free preview version of DALL-E 3 provided by Microsoft's Bing, although @Saracenus is using the fancy paid version through his OpenAI account.

 

[Saracenus]

How are you getting all this to look right? Every time I've asked DALL-E for art (we have an account through work) I get misaligned weapons (like bows with two or three arrows nocked), strange numbers of fingers, limb misalignment and other oddities.

Well, there is plenty of that in what I produce, though it less than it was before. Generally I have to go through a couple of prompt evolutions to dial in on what I want. I have also put stuff in Photoshop and used it powerful suite of tools now augmented by Generative AI to remove artifacts and alter images to get them just right.

I do not know if you are using a full on OpenAI.com account with GPT4 and DALL-E 3, but I have a paid account I am using that allows me to created dedicated GPT Assistants that help focus my efforts on creative ideas, rather than technical details. I have a whole post on Tangents, when an object (e.g. Swords, Spears, etc.) is split in two or more parts because it is behind the subject and how to fix that in your prompt. You still get some goofiness but it way less. Also I have another post here on bowstrings and archery stances.

I cannot prove that it is remembering previous chats, but since I fed a bunch of data to my GPT Assistant on archery for that specific chat, the images with bows in later chats has remarkably improved. I hope this helps.