?Don't mention her name in your prompts otherwise data about her already in your model will disturb the outcomes. I actually made this lora just for avoiding bad performance of her outputs in original SD v1.5. Just type in words like "woman" (you could refer to my png info) and use this lora. ~~ Trained a week ago and has been tested many times. Weighted 0.8~0.9 might be good. \@o@/ Hires.fix on is suggested as always (e.g. 448*640--1.6x-->716*1024, Hires steps 18,Denoising strength 0.45, latent.) ~~ Be kind to these lovely. Legal & RiskIt is prohibited to use this model for commercial purposes and any scenarios of illegal acts and purposes. |