A LORA to help with the O-Face
Description
FAQ
Comments (5)
Looks nice but does it have trigger words that work better than others, or does it just enhance existing tags? Usage is a bit vague from the examples.
There are no trigger words, it's mainly down to the strength to control and enhance the existing prompt. The dataset included these but they aren't essential: orgasm, beautiful agony, face focus, view from above, orgasm expression, face, shoulders, hair, breasts.
almost 900mb?
i feel like that could be trimmed
like dude, HDD storage costs like $0.014 per GB and even PCIe 4 NVMe drives are like $0.05 per GB, if the model works well and is not over sampled who cares. even if every lora were this size and you had 100 loras thats only 90GB, or $4.5 worth of NVMe storage.
The expressions are alright, although what I'm generating is much less pronounced than the sample images.
The extra learning is an issue. I'm doing half-portrait images (above the waist). Even with "semen", "squirting", "ejaculate" and "fluids" as negatives, pretty much every images either has fluid shooting up from out of frame, or semen on her body (occasionally there was drool/post-blowjob semen). Also has the standard issue for adult Lora of pushing breasts towards F-cups.
As it is, can't recommend this. It'll work in some situations, but there's a lot more to it than advertised.