HideShow Thots
one of the more frustrating things for me when it comes to generating pretty ladies is the clothing. i don't really want to define the clothing but if i don't i end up with a lot of naked ladies and that's not what i'm after. so i've mostly settled on using 'lingerie' in my prompts but that still often gives topless ladies so i need to add 'bra' to the prompt as well! i've also tried variations on bathing suit, swimming suit, etc. but find the prompts for underwear to be more in line with what i'm after.
in this batch i was using lists of wildacrds as variables via the stable-diffusion-webui-wildcards extension. __haircolors__ and __era__ specifically. the redheads are just ridiculous here, almost always look liek a fresh dyejob or clown wig so... yeah not happy with them that but otherwise i think this is a pretty good batch. i've done no cherry-picking or culling or deleting of anything, this is every image generated by the prompt in this run.
i'm not necessarily after perfection, just something that looks realistic for the most part. different people have different ideas of what "perfect" and "realistic" are so... yeah i might not care what someone else thinks.
keywords like 'hair' and 'eyes' are in there in an attempt to prevent the heads from getting cut off or cropepd out. 'hips' to try to get more of a body shot than a bust-up portrait. '(feet:02)' to hopefully get a full body shot, the ':0.2' is a reduced attention/emphasis weight. Stable Diffusion is notroiously bad at hands and feet (i usually have hand and finger related keywords in my negative prompts because of this) so i don't want it giving me any closeups of horribly wrong looking feet if i can avoid that.
age related keywords in your positive and negative prompts are also really important for reasons obvious to me.