It's not that it can't be generated, but the quality is pretty questionable.

-If you generate it with a higher step size, it will be somewhat better.

-But for some reason, when the strength is high, it splits.

-Images tend to be generated from a high angle.

-It seems the model isn't learning “shuriken” properly. If I remove “shuriken” from the prompt, the art style becomes reasonably consistent.

Prompt:

tail, japanese clothes, mask, no mouth, mask on nose, scarf, orenge clothes, forehead protector, greave, toeless legwear