WHAT CHANGES IN VERSION 2.

I found a fact that if you over highres the dataset, you'll get horrible resuilt when using hires fix when generating. I'm 100% SURE that this one would get a MUCH BETTER result than the previous version 1.

FOR PERSONAL USE ONLY. WILL BE DELETED IF REQUESTED.
If you have the means, please also support the original creators.

Feel free to DM me Character/Outift suggestions. Tips via Ko-fi or Civitai Buzz are appreciated.

https://ko-fi.com/rggg716871 *

Recommended Appearance/Outfit (I just used a little images for each outfit, DO NOT expect the outfits results are perfect)

CHECK EACH OUTFITS IN SAMPLE POST

CLASSIC

da7wata, small breasts, blonde hair, brown eyes, thick eyebrows, long hair, blunt bangs, eyelashes, pearl earrings, serafuku, blue sailor collar, green neckerchief, puffy short sleeves, blue pleated skirt,

HIGH PONYTAIL

da7wata, small breasts, blonde hair, brown eyes, thick eyebrows, high ponytail, long hair, blunt bangs, eyelashes, pearl earrings, serafuku, blue sailor collar, green neckerchief, puffy short sleeves, blue pleated skirt,

CONVENIENT STORE

da7wata, small breasts, blonde hair, brown eyes, thick eyebrows, short messy bangs, long hair, blunt bangs, eyelashes, pearl earrings, white shirt, wide sleeves, short sleeves, midriff peek, jeans, baggy pants,

EXTRA (NOT TRAINED)

o-ring choker, white choker, serafuku, (taut shirt:1.1), (cleavage:1.1), (collarbone:1.1), covered nipples, (scoop neck:1.1), blue sailor collar, (bare shoulders:1.1), green neckerchief, crop top, midriff, blue pleated skirt, microskirt, highleg panties, side-tie panties, (see-through legwear:1.1), white legwear, (kneehighs:0.7),

A LoRA strength of 1.0 ensures maximum outfit accuracy but impacts the overall art style (character features remain intact). Adjust the weight as needed. (The sample is .9)

Setting

• Checkpoint: WAI-illustrious-SDXL v16

• Lora: Noob Hyper Dmd, strength: 0.55

• Lora: T85P55 (Optional, for better head-to-body ratio)

• Sampler name: euler a | Scheduler: normal | Cfg: 1.2 | Steps: ~12

• HighRes-Fix: 4xUltrasharp | x1.25 | Steps: 12 | Denoise: 0.35 | Iterations: 1~2 (Optional, Depends on your composition)

• (If you're not using Noob Hyper Dmd, just use a general metrics for generation)

DISCLAIMER:
THIS LORA MODEL IS CREATED SOLELY FOR ACADEMIC RESEARCH AND PERSONAL USE. IT IS STRICTLY PROHIBITED TO USE IT FOR COMMERCIAL PURPOSES OR PROFIT. ADDITIONALLY, IT MUST NOT BE USED TO GENERATE NSFW CONTENT. ANY CONSEQUENCES ARISING FROM VIOLATIONS OF THESE TERMS SHALL BE THE SOLE RESPONSIBILITY OF THE USER.