Hello it's OKGC, kpop girls/actress lora maker. I speak Mandarin, English, and a little Korean. Follow the info below to get better result. Feel free to ask for help.
Recommended checkpoints:
supashy mix - https://civarchive.com/models/156202/supashymix
final version works best (what I'm using), get it on supashy's gumroad to support him please - https://supashy.gumroad.com/l/supashy_mix_final?layout=profile
Moonfilm Reality 3.0 - https://civarchive.com/models/43977?modelVersionId=73071
majicMIX realistic v7 - https://civarchive.com/models/43331?modelVersionId=176425
Recommended parameters: (works best in my opinion)
Trigger word: baejuhyun
Sampling Method: Euler
Steps: 50
Hires. fix Upscaler: 4x_NMKD-Superscale-SP_178000_G - https://huggingface.co/gemasai/4x_NMKD-Superscale-SP_178000_G/tree/main
Put it in folder models/ESRGAN
Hires upscale: 2
Hires steps: 15
Denoising Strength (Hires): 0.25
Clip skip: 2
CFG scale: 7
Size: 512x768
vae: vae-ft-mse-840000-ema-pruned - https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.ckpt
Put it in folder models/Vae
Eta noise seed delta (ENSD: 31337)
Find it in SD webui/Settings
If the face comes out badly, use adetailer to fix it.
https://github.com/Bing-su/adetailer
Recommended parameters: (works best in my opinion)
1st Adetailer model: face_yolov8n.pt
Inpaint denoising strength: 0.25
2nd Adetailer model: mediapipe_face_mesh_eyes_only
Inpaint denoising strength: 0.25
If the face detection of adetailer detects the face of other people in the background,
set Detection model confidence thereshold: 0.8~0.85
Used embeddings:
(how to use?: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Textual-Inversion)
ng_deepnegative_v1_75t - https://civarchive.com/models/4629/deep-negative-v1x
badhandv4 - https://civarchive.com/models/16993?modelVersionId=20068 (add disfigured in negative prompt may also helps getting better hand)
bad_pictures - https://civarchive.com/models/23178?modelVersionId=66043
Highly recommended tools: Lora Block Weight - https://github.com/hako-mikan/sd-webui-lora-block-weight
How to use?
<lora:lora name:1:lbw=IN02>
<lora:lora name:1:lbw=1,1,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0>
Recommended parameters (name whatever trigger word you want, it's all about the block weight)
Body:
WINGS:1,0,0,0,0,1,0,1,1,1,0,0,0,0,0,0,0
BODY:1,1,1,1,1,1,1,1,0,0,0,1,1,1,1,1,1
BODY0.5:1,1,1,1,1,1,0.2,1,0.2,0,0,0.8,1,1,1,1,1
Face: (face shape, hair style, eye shape, eye colors, etc.)
EYE:1,0,0,0,0,0,0,0,0,1,1,0,0,0,0,0,0
MOUTH:1,0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0
FACE:1,0,0,0,0,0,0,0,1,1,1,0,0,0,0,0,0 (which I use for my lora)
FACE0.5:1,0,0,0,0,0,0,0,0.8,1,1,0.2,0,0,0,0,0
FACE0.2:1,0,0,0,0,0,0,0,0.2,0.6,0.8,0.2,0,0,0,0,0
Hands:
HAND:1,0,1,1,0.2,0,0,0,0,0,0,0,0,0,0,0,0
outfit: (with tags)
CLOTHING:1,1,1,1,1,0,0.2,0,0.8,1,1,0.2,0,0,0,0,0
pose: (with tags)
POSE:1,0,0,0,0,0,0.2,1,1,1,0,0,0,0,0,0,0
palette/color style: (with tags)
PALETTE:1,0,0,0,0,0,0,0,0,0,0,0.8,1,1,1,1,1
Character:
KEEPCHAR:1,1,1,1,1,0,0,0,1,1,1,1,1,1,1,0,0
Background:
KEEPBG:1,1,1,1,1,1,0.2,1,0.2,0,0,0.8,1,1,1,0,0
Reduce overfitting (same as OUTALL）
REDUCEFIT:1,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1