Update:
Pony. Should actually work consistently on Pony and AutismMix now. Apparently already works well enough on landscape
Update for XL v1.0:
oh yeah XL is a thing
also chairs
also recommended weight still 1.0
Update for v3.5:
nothing much except made the guy actually show up in the frame more consistently
also chairs
Update for v3.0:
should have good character and style LoRA compatibility all around now
currently tailored to images of width 512 of only (512x768 etc)
adding the tags girl on top and straddling appears to improve generation consistency somewhat, but I haven’t done extensive tests on it
Probably better to use weight 1.0 for this one
Previous versions:
tested on anything v4.5, corneosu7thheavenmix, and AOM3
recommended weights: 0.7-0.8
compatibility with character LoRAs: case by case basis. Just put the intended weight of whichever of them for the best chance that it works out
optional tag for hugging and more visible genitals
Description
hopefully reduced overburnt images and increased compatibility with character and style LoRAs due to being trained from a bigger dataset this time.
FAQ
Comments (9)
Does it works with realistics models ?
chilloutmix seems to yield decent results
@TheArnoldPickaxe ok thks. I'll try soon :)
The real-life model doesn't seem to integrate well, but using plugins can slightly mitigate this. The model training is good, but the body gets twisted due to other LORAs! Still, I'm giving you 5 stars, well done!"
Good
Could you please make a LoRA for piledriver position? :)
Why do some of these LORAs mess up faces so bad?
Lack of diversity in the dataset and/or overtraining the model.
很可惜，只能用在2D上面，3D会强制应用2D效果，效果很差