"The year on this planet equals 20 years on the Earth".
"Great. That where we are going to finish the development of 'School of Lust' "
*insert meme picture here*
Chill af.
I'm not even sure of the fact...
Add POV anal doggy for Flora and we're golden. There's a serious shortage of such model-generated pictures - not the shitty ones with graphical artifacts or not the ones with non-required exaggerated features from other \ extreme fetishes.
I presume this is because the image-generating-models have been:
- "trained" upon a shitty datasets with poor annotation
- or the sample of such annotated pictures in the "train" dataset have been really small
- or there was a sample that is extremely biased towards other co-existing fetishes with them on the one picture
From the such pictures I've seen there always seem to be a lot of "sweating" \ "oiled" \ "sperm drops" artifacts on a females asses on a picture. It seems that the resulting image-generated model simply can't separate "anal sex" from "sperm drops on the ass", for example. Not everyone's piece of cake.
I'd say it's a general problem with NSFW-pictures that people try to make in hopes of easy money. Another example - POV pictures are over-represented with the vaginal sex overall in any sample datasets, so... all sorts of weird graphical artifacts appear in the output when people ask the model to generate it with anal sex - I presume that's why there are so few of them.
Also, if you'll manage to make any serious improvements of overall consistency of characters looks across the different pictures (this is a fundamental problem with the image-generating models) - you can even re-brand and make your own game. Just like it's been with Urban Demons some years ago.
