v2

finally time to upgrade

use this as when you generating "masterpiece,best quality,batman,solo,grey bodysuit,utility belt,black cape,black gloves,superhero"

then For best results, in the end add short prompts like "eating pizza" or "squatting"

sample eualer a and steps 30-35 works best

all images generated weight 0.7-0.8

ı also use vae-ft-mse-840000-ema-pruned

v1

edit : ı fell like this prompt works best and it does keep design more consistent without white skin stubborn appearing"batman,solo,grey bodysuit,utility belt,black cape,black gloves,superhero"

made this as a fun experiment to see if it's possible to create a Batman character Lora using Midjourney which does pretty good job of replicating batman design accurately

Trigger Word is "batman"

most of my training images created with prompt" batman comic book style, 90' s comicbookstyle"

recommended weight 0.5-0.6 it is not bad for what it is, but definitely needs version 2

maybe some day ı could do another one with much bigger datasets, until then ı hope you like it