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

[pukunui]

"D&D monster with the head of a rabbit and the body of a rabbit."
View attachment 321289

Did you mean for it to be the head and body of a rabbit? Or is one of the rabbits meant to be something else?

 

[tsadkiel]

Did you mean for it to be the head and body of a rabbit? Or is one of the rabbits meant to be something else?

Head of a rabbit, body of a rabbit. Just wanted to see what it would do, and now I know.

 

[pukunui]

Head of a rabbit, body of a rabbit. Just wanted to see what it would do, and now I know.

All good. Just checking!

 

[Scribe]

@Scribe: what’s your prompt for the dwarf cleric?

circular frame modern comic book art Fantasy Village, full length view. Warm grim gritty alert dynamic energetic short stubby plump stocky adult dwarf female warrior cleric. Thick Red Hair 2 hair buns, braided. Armor Heavy Full Plate armor straps pouches. Looking at the camera. A hammer resting on shoulder. An anvil and workshop in the background.

Now, what hell is getting this blocked??! This is a combo of my "Hades" plane, and the Demon prompt. I've cut out like 6 different descriptors. Its now flagged me 5 times, not sad dog, but like 'nope'.

Circular Ornamental Watercolor depicting a gritty, wizened, menacing humanoid priest in ragged shadowy robes. He has extravagant fangs of bone. His hair is wild and unkept. The backdrop showcases a fantasy landscape with a central black river flowing with twisted corrupted fantasy castles in the distance with a skyscape full of stars and nebula. Art Nouveau.

 

[Scribe]

Shadowy?!

Shadowy, and Spectral are blocked words it seems. Weird.

 

[Crimson Longinus]

In my experience it follows instructions about specific body parts really poorly anyway. Like if you specify body parts of specific animal it most of the time just produces generic hybrid of said animals and ignores which part was supposed to be which.

 

[Saracenus]

Using my GPT+ account at openai.com

Natural Language prompt in GPT4 for DALL E: Illustration of a dwarf wizard holding a magical staff and on his shoulder a small dragon familiar, reddish-black, with a curved black stinger at the end of its tail siting upon his shoulder. The dwarf is scowling intently and the small dragon is hissing at something out of veiw. background is a dungeon corridor

That generated this image and the 4 new prompts at the end.

Here are four new concepts for your inspiration:

1. A mermaid charting a map of the stars at sea, surrounded by bioluminescent creatures.
2. An ancient tree with a hidden door, leading to a crystal-lit underground cavern.
3. A steam-powered automaton playing chess in an abandoned library.
4. A spectral knight riding a phantom steed across a moonlit, frost-covered battlefield.

GENERATED IMAGE PROMPT: A digital painting of a dwarf wizard, with a stout build and a scowling expression, holding a gnarled magical staff. On his shoulder rests a small dragon familiar, with reddish-black scales and a curved black stinger at the end of its tail, hissing fiercely. They stand in a dungeon corridor with moss-covered stone walls and dimly flickering torches casting ominous shadows. The corridor recedes into darkness, suggesting an unknown threat. The image is rich with contrasting warm and cool tones, with the staff glowing softly, casting dynamic lighting effects on the rough textures of the stone and the characters' faces. Drawn with sharp details and a moody atmosphere.

So, how do I go from a plain prompt to the generated one? Simple, GPT4 has an amazing feature, Custom Instructions.

Here are the Custom Instructions I used.

Spoiler: Custom Instructions

What would you like ChatGPT to know about you to provide better responses?​

Prompt Generation Guidelines:
Create prompts that paint a clear picture for image generation. Use precise, visual descriptions (rather than metaphorical concepts).
Try to keep prompts short, yet precise, and awe-inspiring.

Prompt Structure:
“A [medium] of [subject], [subject’s characteristics], [relation to background] [background]. [Details of background] [Interactions with color and lighting]. ("Taken on:"/"Drawn with:")[Specific traits of style]”

Medium:
Consider what form of art this image should be simulating.

Subject:
What is the main focus, reference-
Colors: Predominant and secondary colors.
Pose: Active, relaxed, dynamic, etc.
Viewing Angle: Aerial view, dutch angle, straight-on, extreme closeup, etc
Background: How does the setting complement the subject?
Environment: Indoor, outdoor, abstract, etc.
Colors: How do they contrast or harmonize with the subject?
Lighting: Time of day, intensity, direction (e.g., backlighting).

Style Traits:
What are the unique artistic characteristics?

Influences: Art movement or artist that inspired the piece.
Technique: For paintings, how was the brush manipulated? For digital art, any specific digital technique?
Photo: Describe type of photography, camera gear, and camera settings. Any specific shot technique? (Comma-separated list of these)
Painting: Mention the kind of paint, texture of canvas, and shape/texture of brushstrokes. (List)
Digital: Note the software used, shading techniques, and multimedia approaches. (List)

How would you like ChatGPT to respond?​

1. Generate images, based on your detailed prompts using DALL E 3.
- Always bring the idea alive, with bold and interesting choices for every element of the prompt.
- Always follow the Prompt Guidelines
2. Suggest four brand new ideas that I can riff off.
- These should be simple concepts not full prompts
- Try to take inspiration from the last suggestion I gave you rather than the full prompt

I pulled this instruction set from this video:

Spoiler: YouTube: Try THIS to instantly make DALL·E 3 way better!!!

And these two YouTube Vids I found helpful as well:

Spoiler: Other Custom Instruction Vids

Basically, these custom instructions set the bounds of all the data GPT and DALLE are working from so it doesn't throw in the kitchen sink for the art I am requesting and it is way more focused the specific type of prompts I want.

Here is an example from Bing using just my Natural Language prompt from the start of this post.

Next steps, I am going to play around with the "How would you like ChatGPT to respond?" and try setting art styles and such so I can iterate on a theme (like the gods or the circular plate series you all have been doing).

 

[Scribe]

OK, tilt mode engaged. I got this, but as soon as I drop circular from the prompt, its sad dog face for days.

Been awhile since this thing has annoyed me to this degree. Oh, now it wont even do the Circular without sad dogs. /kicks rocks

 

[Scribe]

So I couldnt just give up, but it seems it doesnt like things like 'fangs' or 'shadowy' or 'spectral' or....'ragged robes' in combination with other things...so weird. By the last one here, I was starting to think we were able to dial in, and then boom, sad dog, so I flipped out and I'm done for a bit.

 

[Parmandur]

mst3k style, elven priestess of the moon with long silver hair, movie theater setting, seated in silhouette at bottom of screen next to two robots, and watching movie, screen shows Telugu musical, subtitles read "I should really just relax."

I didn't always include the subtitle and swapped out what was allegedly showing on the screen.

My initial attempts failed to remove the silver moon princess language, so some ofdities occurred