Old wildcard prompt

So I was like "Hey, I really love that realCartoon3d checkpoint, let's try out that new prompt!". Only I didn't use the new prompt, I used the old one I was trying to replace.

I ran a 20x2 grid (40 images) and accepted 17 of them. I used to review them in Faststone Image Viewer, but I started using the free Adobe Bridge, which has a review function that lets you reject pictures. This made it really easy for me to indicate duds.

Generally, I don't really care about the prompt. But my objective here was to get a prompt that yields images that adhere to it better.

My old one wasn't very discerning about clothing, so you could specify multiple conflicting outfits.

Here's the prompt I'm intending to replace:

photo of 
{2::a __nsp/nsp-adj/adj-beauty__ __nsp/nsp-body/body-fit__ __jumbo/people/nationalities/all-nationalities__ woman| 1::__personal/girls__},
wearing {1-3$$ and $$__devilkkw/attire/*__},
with hair styled as __hairstyles/female/*__,
BREAK ({__personal/locations__|in the __jumbo/places/*__}:1.1),
__promptgeek/lighting__, __promptgeek/angle__, __promptgeek/framing__, __promptgeek/photorealistic/camera/properties/shot_on_digital_camera__

As you can see, this one incorporates the prompt geek guidance for adding variation using lighting, angle, framing, and camera. Most of the time it will give you a girl with a random ethnicity, and sometimes it will use embeddings and loras to give you a celebrity.

I made two changes to this in my updated prompt:

1. I removed the conflicts that can occur with the attire by getting more granular and making them mutually exclusive

2. I moved some of the prompt geek settings to higher up on the food chain, because they weren't taking effect so far down

Most of the images that I nixed were simply due to bad hands or other aberrations. Some of them I simply didn't think met muster. Less than half were hits, and of those, there was generally significant deviation from the prompt. It's my hope that my new prompt will perform better in this regard. I'll be running the batch in 2:3 rather than 1:1; that might also have an impact as well, given the nature of the subject.