First LoRA trained on Civit

Go give the artist all your money. @HaruFeng on Twitter

All example images generated with AutismMix_confetti with Invoke, naively upscaled x2, no edits and no extra LoRAs to best demonstrate this model. Better results can definitely be achieved with better upscale/hires fix methods and a little inpainting (adetailer etc)

All prompts are some variation of:

score_9 score_8_up score_7_up source_anime haru feng
1girl etc
hair style etc
body type etc
clothing etc
background etc
hair colour, eye colour