And for the final Phantasia girl, at least for the time being.
Wish I could generate better fighting poses with her, maybe I should move away from Pony?
そして、少なくとも今のところ、最後のファンタジアの少女。すずともっといい戦闘ポーズをとれたらいいのに。Ponyから離れるべきかな?
For the prompts
suzuBase, brown hair, brown eyes, long hair, high ponytail,light red suit, ninja, yellow scarf, fingerless gloves, There are a few other costumes, but since their data were so limited, they may not be coming out very well, you can refer to my sample images.
他にも衣装はいくつかありますが、データが少ないので、うまく出ていないかもしれません。
生成したい場合は、サンプル画像を参照してください。
Description
FAQ
Comments (3)
I think the issue with poses is how the majority of training data doesn't have her full body, ending at the knees and the tags given don't really give anything to extrapolate (it's tagged as a onesie instead of shirt and pants going by the trigger words). Seems to work fine if pose is only upper body/cowboy shot, but even static "full body" tends to get weird at bottom.
I think there was a little above 1/4 of the data were full body, it was quite hard to think of tags for the ninja suit, so maybe the limitation of tags? Still Pony probably doesn't have much data with katana wielding, which makes it quite hard to generate them.
@GanGanGan Oh, that's your problem. SD in general sucks at holding objects, especially without a LoRA dedicated to that one object.
As for tags, I'd say "japanese clothes, yellow scarf, red dougi, black belt, red baggy pants, fishnet gloves, fingerless gloves, red gloves, red moccasins, red shoes" (there's likely a better name for her footwear)
Edit: Maybe make an article (it's free) asking for tag recommendations if you're stumped on a character's outfit in the future?
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.