D&D General AI Art for D&D: Experiments

Parmandur · Nov 25, 2023

1950's technicolor hollywood sci-fi film of a kobold running through an underground passage by torchlight. Kobolds are small, evil dog-like men usually live underground. They have scaly rust-brown skin and no hair.

pukunui · Nov 25, 2023

There were some pretty successful dragonborn earlier in the thread. I think some of them were made using "lizardman".

mamba said:
humanoids with scales and dragon heads in armor walking on forest path, high detail digital painting

View attachment 299234
View attachment 299235
View attachment 299236
View attachment 299237

Scribe said:
As I prepare to put my headphones on and perform some angry work, I wanted to do a quick palate cleanse.

This is almost exactly what I wanted, very happy with it. Dragonborn Gish.

View attachment 305666

@EzekielRaiden as the resident Dragonborn aficionado how important is the tail to you.

Scribe said:
My Sorcerer, but as a Dragonborn.

View attachment 321270

View attachment 321271

Saracenus said:
Here is a series of a gold dragonborn Regency era aristocrat at a ball... Detailed Digital Art...

View attachment 317944 View attachment 317945 View attachment 317946 View attachment 317947

View attachment 317948 View attachment 317949

Here is the next series with him holding a ornate cane...

View attachment 317950 View attachment 317951 View attachment 317952
View attachment 317953 View attachment 317954

Parmandur · Nov 25, 2023

1950's technicolor hollywood sci-fi film of a kobold in a dimly lit cavern carryin a spear. Kobolds are small, evil dog-like men usually live underground, they have scaly rust-brown skin and no hair.

Parmandur · Nov 25, 2023

pukunui said:
There were some pretty successful dragonborn earlier in the thread. I think some of them were made using "lizardman".

Ah, thanks I was trying to engineer it from scratch.

First result playing off of these is...not Dragonborn, but interesting.

Andvari · Nov 25, 2023

Spent way too much time yesterday generating some fantasy scenes.

Can any imaging service yet handle prompts to change a specific part of an image? That would be quite an important feature as you can then take an image you are overall satisfied with and correct specific mistakes, rather than having to generate a completely new image and hope for the best.

pukunui · Nov 25, 2023

Some different takes on Savil, the half-elven mayor of Dockalong in "Wedding Bells":

I'm using "tall halfling" rather than "half-elf". Trying for something a little more matronly. A lot of the ones DALL-E's been spitting out are a bit too young looking. I also feel like it's quality is a bit low at the moment. More weirdness with hands and eyes and such than usual.

I used up all of my 15 boosts for the day (and then some), and the ones I've shared above are the best of the bunch. Can't really decide which one I like best either.

EDIT: Here we go! I think this is the (slightly 'shopped) winner!

Yora · Nov 25, 2023

I was wondering what a Bloodborne boss monster drawn by Larry Elmore would look like, since I'm working on a campaign idea that is basically "Bloodborne meets 90s AD&D".
Out of curiosity, I only entered "Bloodborne great old one in the style of Larry Elmore" and nothing else to see what it does with that.

It's not a squid, but actually really not bad. I might try to use that red guy.

Saracenus · Nov 25, 2023

Blue Orange said:
It threatened me with a term of service violation for 'fauvist', perhaps confusing it with a certain ideology.

I am assuming that was Bing? OpenAI didn't blink...

Saracenus · Nov 25, 2023

Tangency... the bane of my existence, turning an otherwise amazing image into one that I cannot use... I finally asked my GPT Assistant on what to do about it...

First, an example of Tangency... The sword behind the back of the Knight does not line up...

ME:
AI art has a problem with Tangency, how can I solve this with you in my prompts.

Fantasy Artisan (FA):
To address the issue of tangency in AI-generated art, it's essential to be specific in your prompts about the placement and interaction of objects in the scene. Tangency in art refers to the problem where two or more elements touch in a way that flattens the sense of depth or creates an awkward visual intersection. This can make the composition less dynamic or visually confusing. Here are some strategies to mitigate tangency in your prompts:

Specify Clear Overlaps: Indicate which elements should clearly overlap others. For example, instead of saying "a bird flying near a tree," specify "a bird flying in front of a tree, with its wing extending beyond the tree's outline."

Avoid Ambiguous Intersections: Be precise about where objects are positioned relative to each other. For instance, instead of "a cat sitting near a flower pot," you could say "a cat sitting behind a flower pot, with the pot partially obscuring the view of the cat."

Describe Depth and Distance: Include details about the spatial relationship between objects. For example, "a row of trees, each progressively smaller and further away, creating a sense of depth."

Use Angles and Perspectives: Specify angles that avoid tangency. For example, instead of "a person holding a staff," try "a person holding a staff at a 45-degree angle, ensuring it doesn't align with the arm or leg lines."

Mention Lighting and Shadows: Use lighting and shadows to separate elements. For example, "a lamp casting a shadow on the wall, with the shadow clearly separated from the lamp's base."

Emphasize Foreground and Background: Clearly define what is in the foreground and what is in the background. For example, "a boat in the foreground with clear water around it, and mountains far in the background."

Create Dynamic Compositions: Suggest dynamic compositions that naturally avoid tangency. For example, "birds flying in a scattered formation, each at a different distance and angle to avoid overlapping."

By integrating these strategies into your prompts, you can help guide the AI to generate images with more depth and clarity, reducing issues with tangency.

ME:
When you, the AI, place and object, example: a sword, behind a subject, the two elements of the object shown do not visually correspond. How would I specify that they should?

FA:
To ensure that elements like a sword positioned behind a subject visually correspond and align correctly in AI-generated art, you can employ specific descriptive techniques in your prompts. These techniques help in guiding the AI to understand the spatial and perspective relationships between different elements, such as parts of an object that are partially obscured. Here are some tips:

Specify the Continuity of the Object: Clearly describe how the object should look as it extends from the visible part to the obscured part. For instance, "a long sword held by the subject, with the visible part of the blade extending straight behind the subject's head, suggesting the continuation of the same straight line."

Describe Relative Positioning: Indicate the position of the object relative to the subject. For example, "a sword slung across the subject's back, with the hilt visible over the right shoulder and the tip emerging from behind the left hip, maintaining a diagonal alignment."

Use Perspective and Angle Descriptions: Provide details on the angle at which the object is held or positioned. For example, "the subject holding a sword angled upwards at 45 degrees, with the lower half visible in front and the upper half extending behind, parallel to the subject's spine."

Detail Proportions and Size: Describe the size of the visible part of the object and how it should proportionally correspond to the obscured part. For example, "a large broadsword with two-thirds of its length visible in front of the subject and the remaining third extending behind, proportionally consistent."

Mention Symmetry or Asymmetry: If the object has symmetrical features, mention how they should align. For example, "a double-edged sword, with both edges visible in front and continuing in alignment behind the subject."

Highlight Shadow and Light Effects: Describe how shadows or light reflect off the object to give clues about its shape and position. For example, "the sunlight glinting off the visible part of the blade and casting a corresponding shadow that suggests the continuation of the blade behind."

Use Analogies or Comparisons: Compare the positioning to a familiar object or scenario. For example, "like a flagpole carried on the shoulder, with the sword's handle and part of the blade visible in front and the rest extending back."

By incorporating these detailed descriptions into your prompts, you help the AI better understand the spatial relationships and ensure more accurate and realistic positioning of objects in the generated artwork.

Using this advice I modified the prompt and referenced the Generation ID to create this...

GENERATED PROMPT FOR IMAGE 1: A painting of a human woman in full plate armor, standing heroically, holding a glowing sword aloft. She is portrayed in a Fauvist art style, characterized by bold colors and strong forms. The armor is intricately designed, shimmering with metallic hues under a dramatic sky. The glowing sword casts a radiant light, creating a striking contrast against the twilight background. The scene is set outdoors with a hint of a rugged landscape in the background, emphasizing her formidable presence. The lighting is dynamic, with the sword's glow being the primary light source, casting dramatic shadows and highlighting the armor's details.
GENERATION ID: 3VbTJx2QhopJNC74

WHAT I ASKED IT FOR: Using Generation ID 3VbTJx2QhopJNC74 as a reference, visualize the following modified prompt: A painting of a human woman in full plate armor, standing heroically, holding a glowing sword aloft with it angling down straight behind her back, suggesting the continuation of the same straight angle. She is portrayed in a Fauvist art style, characterized by bold colors and strong forms. The armor is intricately designed, shimmering with metallic hues under a dramatic sky. The glowing sword casts a radiant light, creating a striking contrast against the twilight background. The scene is set outdoors with a hint of a rugged landscape in the background, emphasizing her formidable presence. The lighting is dynamic, with the sword's glow being the primary light source, backlighting the subject by casting dramatic shadows and highlighting the armor's details.

GENERATED PROMPT FOR IMAGE 2: A painting of a human woman in full plate armor, standing heroically, holding a glowing sword aloft with it angling down straight behind her back, suggesting the continuation of the same straight angle. She is portrayed in a Fauvist art style, characterized by bold colors and strong forms. The armor is intricately designed, shimmering with metallic hues under a dramatic sky. The glowing sword casts a radiant light, creating a striking contrast against the twilight background. The scene is set outdoors with a hint of a rugged landscape in the background, emphasizing her formidable presence. The lighting is dynamic, with the sword's glow being the primary light source, backlighting the subject by casting dramatic shadows and highlighting the armor's details.
GENERATION ID: XB41dAjzMmEFFJCS

This is not perfect. I got lucky with the first image created actually followed the prompt exactly... here are the next four generations of the same prompt...

Saracenus · Nov 25, 2023

Just as control, I started a brand new chat in Fantasy Artisan and put the following prompt in:

PROMPT: A painting of a human woman in full plate armor, standing heroically, holding a glowing sword aloft with it angling down straight behind her back, suggesting the continuation of the same straight angle. She is portrayed in a Fauvist art style, characterized by bold colors and strong forms. The armor is intricately designed, shimmering with metallic hues under a dramatic sky. The glowing sword casts a radiant light, creating a striking contrast against the twilight background. The scene is set outdoors with a hint of a rugged landscape in the background, emphasizing her formidable presence. The lighting is dynamic, with the sword's glow being the primary light source, backlighting the subject by casting dramatic shadows and highlighting the armor's details.

That generated the following: