Just a model i created for fun for testing sd xl lora :D a bit jank without negative emmbeding and even more jank without any at strength 1 (hit or miss at 1 but some what good most of the time)
edit: apparently its really good using anime art diffusion xl using the negative emmbedings and really good in blue pencil aswell aslo the jank output was due to the res being to high opps me :/
Resolutions that i find best are: 915x1144, 886 x 1182, it can do 1024x 1546 (768x1344 works better) but i would not do it (gives the best result at the expense of a chance in jankyness 2/5 chance jank unless using negative emmbedings it eill reduce the chance)
trained on 99 images
stregth try using 0.8-1 works well on blue pencil sd xl lineups (works well on 0.9)
please give feedbacks as i am playing around with a lot of settings and want to know the optimal settings are :DDD
any comments, suggestions, and tips are appreciated :D hope you enjoy it :DDD