Do you use bing as your search engi e, or are you using this on Edge? I wonder if it is using your search history as an "unspoken prompt."
D&D General DALL·E 3 does amazing D&D art
- Thread starter M.T. Black
- Start date
Saracenus
Always In School Gamer
So, in my quest to create a cool version of the Elemental Eye symbol, I tried something new. I fed a natural language description in ChatGPT 4.0 requesting a prompt for Dall-E 3.0. Then I cut and pasted the output into Bing image creator...
Unfortunately, the word grotesque triggered some Not Safe Content tripwire and I had to take that word out. I sanded down my prompt a little more, figured out that if I told Bing to rotate a triangle 180 degrees it would actually make the triangle point down 50% of the time.
In the end to get my final image, I had to Photoshop the "Y" into the triangle...
Dall-E 3.0 "original"
Final photoshoped version...
Unfortunately, the word grotesque triggered some Not Safe Content tripwire and I had to take that word out. I sanded down my prompt a little more, figured out that if I told Bing to rotate a triangle 180 degrees it would actually make the triangle point down 50% of the time.
In the end to get my final image, I had to Photoshop the "Y" into the triangle...
Dall-E 3.0 "original"
Final photoshoped version...
pukunui
Legend
I've had another go. I think the first one does a good job of looking more like a halfling, while the others still look more like half-elves. I think the first one also does the best job of "cynical wary ... thief" as well, but then she's got a foot that's on the wrong leg ... sigh!
All of them are "high detail modern comic book" style.
pukunui
Legend
I didn't actually want that much action. This is the prompt that gave me the first, third and fourth: high detail modern comic book, full length view, white background. Cynical wary smart short curvy robust light-fingered female hobbit halfling thief rogue pickpocket medieval fantasy vest blouse pouches fingerless gloves. She is slinking along a dirty cobblestone alley. A rat watches from atop a barrel in the foreground.
The second one had this slightly different prompt: high detail modern comic book, full length view, white background. Cynical wary smart short curvy robust light-fingered female hobbit halfling thief rogue pickpocket medieval fantasy vest blouse pouches fingerless gloves.
Apparently this is what having your back against the wall looks like:
I like them, but again - more half-elf than halfling. (These were also done in a "high detail modern comic book" style. I guess I'll need to add "full color" or something to get it to stop giving me black-and-white drawings.)
I'm all out of boosts.
pukunui
Legend
Yeah, and it still doesn't always work. I mean, how many of my Elsie pics actually look like a halfling? It also objected to me attempting to show her picking someone's pocket, so ...
Here's an outtake from the same generation as #2 above (so High detail modern comic book, full length view, white background. Cynical wary smart short curvy robust light-fingered female hobbit halfling thief rogue pickpocket medieval fantasy vest blouse pouches fingerless gloves.)
First is the original that it gave me, and then is a version where I've erased the figure on the left and then used Photoshop's generative fill to add in her missing elbow. It did a good job of that (and added those stylish lines), but it was unable to add the rest of her legs in any semblance of the same art style.
She is certainly not a halfling, but she could make for a stylish human or half-elven woman.
Chrome! I have experimented using Bing...
I was just curious if that had something to do with it. I have never gotten a terrain like or mini like output.
