Trained using 280 images, capped from all over the anime episodes. Curated, fully designed. Dataset is provided on the right >>>>>
Honestly, the dataset is the reason why some commissions are priced higher. It is the effort to cap, to sort, to cut and edit data properly, not just some half-assed job. You pay for my effort to edit. This alone took me 2 hours capping, 6 hours selecting proper caps and editing. That's whole 8 hours of editing. YOU PAY FOR THE TIME I SPENT TO MAKE IT PLUS ELECTRICITY I USE TO TRAIN IT.
Recommended setting is 25-30 step, 4-5 CFG, euler a.
Use strength between 0.7 to 1. ADetailer is a must for this model.
Reduce strength if you wanna use with other outfits. Preferably 0.7. Upscaled using 4x-FoolHardy-ReMacri.
This is ONE of the LORA intended to teach people how to make their own DATASET on their own.
THERE ARE NO SHORTCUT FOR GREAT LORA.
You can also use this model at
Description
280 x 3 x 10 = 8400