V2
Emilia From Re:zero character lora
Updated with the primary goal, anime style added anime coloring, anime screencap to prompt
Focused on anime styles and quality of outfit
From testing, face, outfit, eye does well, sometimes grow double hair clip
V1 model had an unique sometimes darker shadow removed
Outfit Sample (add anime trigger words)
shimilia, long hair, grey hair, purple eyes, pointy ears, crown braid, hair flower, hair ribbon, purple ribbon, x hair ornament, white dress, cleavage, bare shoulders, wide sleeves, white thighhighs, thigh boots,wide sleeves, long sleeves, detached sleeves (may vary)
remove white thighhighs, thigh boots, if only doing portraits include, if not
Outfit with more anime (more closer to training data prompting)
shimilia, long hair, grey hair, purple eyes, pointy ears, crown braid, braid, hair flower, white flower, hair ribbon, purple ribbon, x hair ornament, white dress, cleavage, bare shoulders, wide sleeves, white thighhighs, thigh boots,V1
Emilia From Re:zero character lora
Only Used Images From Season 1
Recommended Running Settings:
Highly Recommended to use SDFORGE with txt2img + hi-res (if lower VRAM) using hires fix to txt2img outputs
Hires.fix Upscaler: R-ESRGAN 4x+ Anime6B
Upscale 1.5-2
Hires Steps 20-25
Sampling Steps 25-40
Optional: Adetailer for face detail