Hola PrettyFly4aWiF1,
I didn't use Stable Diffusion -- if you want to create explicit or even just sexy images, you're better off with unfiltered Stable Diffusion than what I'm using. And in my experience, prompts don't carry over well from one system to another, or even from one model to another within the same system.
If you like my results, the key is the process. Basically three steps:
(1) LOTS of grinding on generating base images to get some you like. For me, this means throwing away 20 - 30 or more generated images to get one that's appealing and fits what I'm trying to do.
(2) Post-processing with a GAN tool to fix smile (teeth), may also improve eyes and ears. Various websites offer this for free, just google "gan photo restoration tool." But the GAN tool usually makes the hair less well defined, so now you probably want to combine the two images to get the best of each. That means...
(3) More post-processing to incorporate the improved part of the post-GAN image into the original image, or vice versa. The GAN tool subtly changes the face color, so you'll end up with a slightly visible borderline to get rid of. Multiple ways of doing this, like Photoshop, inpainting, or even "spot fix" in Windows photos app.
Good luck! Important: I am a rank amateur, there are tons of people who know MUCH more than me about this stuff.