>>2852170Always describe the girl the best that you can and let the trainning system do the thinking and think will go.
If you tag manually I reccomend Starting from top to bottom to not forget anything
Hair lenght, hair color, hair detail, hair accessories, eyes, neckwear, bodywear, handwear, bototmwear, legwear, foot, background, face expressions, body position, background accessories
Blonde hair, medium hair, sidelocks
Air hornament, horns, bow. whatever
Blue eyes
Detached collar, pendant, necktie
Dress, chinese clothes
Wirstband, gauntlets
belt, obi
skirt, pants, shorts
tighthighs, socks
sneakers, boots
indoors/outdoors
church/garden/park/forest
smiling/surprised/blushing/tears
squatting, sitting, laying, arms up
Holding X
table, bed, teapot,
Now, this is just an order like any other to not forget stuff and even then sometimes you forget.
Thing is the better you describe the character the better.
Keep specifiing stuff, if the lora doesnt add it, just tell it. For example when training the lora if the character uses fingerless gloves, include the fingerless gloves tag.
If when you generate the character you want them and the lora dont put them, just tell "fingerless gloves" in the prompt.
If you are tagging them authomatically, then dont worry, and trust waifu tagger.
Personally I would reccomend for characters lora not using auto tagger and do it manually. You dont need to describe every image you can do it ONCE (two or three times if she has several costumes) then copypaste it on every single image (of the same costume obviously) and add at the end the details like the position and the background.
For style loras use autotagger do its thing because not gonna manually tag 100 different images, obviously But for characters is fine doing manually