Well the time I upload isn't that good huh...

This is my first ever lora model, you can check out my Pixiv for more previews.

4BEraser - pixiv

My version is focus on the official outfit of C.B, with たろうまる's facial expression & art style. The result is better than other one's version but sometimes it can't recognize the single sleeve of chara.

I don't know what is best, but I do usually use a mix model of Counterfeit & PastelMix, other anime model (Anything, Orange) are compatible.

How to use:

You can input nothing to generate the character any way, but I do recommend you to use following prompts to generate the outfit:

• Base: umamusume, green bow, green eyes, horse ears, horse girl, ahoge, bow, hat bow, hairclip, long hair, horse tail, hair ornament,

• Outfit: (white pants:1.2), midriff, navel, bangs, jacket, belt, flare pants, strapless shirt, mini top hat, green shirt, (single sleeve1.4), crop top, white jacket, arm strap, open clothes, open jacket,

  • It isn't designed for alternative outfits, but you can put the Outfit tag in negative prompts for a try, I have put some examples in previews.

  • If you use base prompt only it will be tempting to generate portrait, if so, please add upper body or full body tag.

  • If you find out that it's focusing on chara's feet, you can try to use Controlnet to limit it.

Settings are what I am using, you can change by your own.

• DPM++ Series(SDE Karras, 2M Karras, etc.)

• about 20 steps, CFG scale 5~7

• LoRA Weight: 0.6-0.8

• If using highres

  • ESRGAN_4X (Not good at portrait)

  • Denoising 0.5-0.6.

  • P.S. LoRA training resolution is 768*512.