Trained with 20 epochs and uses cosplay photos as datasets. This one not too strong LORA as other ones so you can use with a little bit stronger settings by rate of 0.8. Setting it to default ones may be less desirable. May requires vae-ft-840000-ema-pruned.ckpt file to use along with it.

See the images here for prompt examples first before using it. Best to use any other cross attention optimization than xformers that is known for unpredictable and undesired results.