Show more for tags - info
may need to add 2d, or anime, in neg for certain 3d, or realistic, styles due to having more 2d image training or 3D, in NEG to enhance
V3 - General - trained on the characters general art (Flexible) - and "may be better" if you prefer a Fan-art/loose approach
V3 - focused - is an attempt to get a more consistent look without strictness - This "may be better" if you prefer 2d art
PonyXL version may need a XL VAE as a 1.5 VAE may cause issues - Images may be - at or past 1024x1024 or 832x1024 - Sampling steps past 30 - this can work with any model that is PonyXL or was merged/trained from it - realistic, or anime, in prompt may help - DOES NOT WORK WITH 1.5 checkpoints - TW is Susiexl,
for appearance
Susie, black eyes, black hair, blunt bangs, short hair, - wavy mouth, pale skin,
(right eye half closed:1:2), (left eye open1:1), - may increase chance of eyes accuracy if weighted
black eyeliner, bags under eyes, - suggest facial mark, in NEG in some cases
adding cartoon, or toon \(style\), in prompt may help depending on style used, and add no nose,
PonyXL - some of these will have increased effect compared to 1.5
for attire
long black dress with white collar,black long sleeves, - white gloves, or wrist cuffs, - black hair ribbon, black shoes,
suggest white sleeves, white dress, buttons, cape, black hair ribbon, in NEG - this may help in General
Description
have been trained longer - may be better - suggest weights for lora 0.8 or lower