Not sure how many stable diffusion tools you're familiar with but off the top of my head I used controlnet to process the original renders into either depthmaps or lineart, both works well in different situations (or together if your gpu can handle it, mine can't so usually i'll try both and see how they look)
for checkpoints I will stick to ones that does 2.5D well and can give decent enough background/environments like
You must be registered to see the links
(I believe this is what I used for the renders I posted here)
or
You must be registered to see the links
(closer to photoreal)
or
You must be registered to see the links
is really good too but the faces tend to give more Asian looking characters as it's mostly trained on those characters so I'd include Asian, Japanese, Chinese, Korean negative prompts to counter that, or include this Lora if its still too much:
You must be registered to see the links
. Character Loras might help too.
I'd avoid porn checkpoints as they tend to give very bland background and lighting
for loras i kept it fairly simple,
use
You must be registered to see the links
at around 0.6-0.8 for darker/more contrast-y lighting,
use
You must be registered to see the links
for better boobs in general
and
You must be registered to see the links
is my personal go to pussy Lora
then it's just prompt engineering, you try to describe the scene best as you can and hope the controlnet don't fuck up the hands (do use
You must be registered to see the links
and
You must be registered to see the links
in negative prompts)
if you don't understand what all of these means, I'd recommend starts from learning the basics of stable diffusion by looking it up on youtube, good luck!