Trained on high resolution sharp images of amateur female women (real photographs not ai) opening mouth wide, sticking out tongue, showing uvula, displaying teeth, also a few nudes to get complete body coverage.
Faces/composition deviate slightly compared to output without this lora. (To maintain face or have it work better together with other/extra loras, look at less aggressively trained Face Portraits - Detailed Skin )
Lora strength around 0.7 should be good for most resolutions up to 1536px; tip: for gens higher than 1536px, lower strength to around 0.6
Description
FAQ
Comments (6)
This would also fit better into concept category I think.
yeah i wondered, but it does change the composition a bit; can i change it? i am not sure.. anyway i changed it
What tool did you use to train this and what were the settings?
ai-toolkit, but on Low Noise setting, and lower learning rate (0.0004) and more steps (8000), I hoped this would cook it more slowly for details..
I might know a lora that would pair well with this. :-)
lets see it :-)