LoCon of El Condor Pasa (umamusume).

Please install extension on WebUI.

Notice

• Here are two version of LoCon.

• One is resized version using sv_fro method. If you interested in, visit here.

• First one used sv_fro method and param=0.99 with kohya_ss resize_lora.py

• I think it is good trade-off with file size and quality.

• The file size has been reduced from 37.2MB to 22.8MB (It saves 40%!).

• The other is original version of LoCon. No resizing.

• Settings of all preview images are same, only LoCon differs.

Recommend options

• LoCon weight 0.5 ~ 0.65(monk costume needs higher)

• Trigger words

  • base

    • el condor pasa \(umamusume\)

    • The mask is not included!

  • mask

    • domino mask

    • Not tagging mask makes her embarrassing. (really)

  • Racing competition costume :

    • domino mask, sailor collar, long sleeves, yellow shirt, buttons, red coat, open coat, belt, pleated skirt, blue skirt, miniskirt, black thighhighs, red footwear, high heels

    • (optional) animal on arm, eagle(Mambo, actually a hawk)

  • Kukulkan Monk :

    • official alternate costume, domino mask, ear ornament, ponytail, bare shoulders, detached sleeves, fur-trimmed sleeves, fur-trimmed shirt, clothing cutout, black gloves, cleavage cutout, black pantyhose, thigh boots, black bodysuit

    • Beware, monk costume needs both higher LoRA Weight and lower CFG Scale!!!

  • Casual clothes :

    • domino mask, yellow shirt, wristband, short sleeves, collared shirt, pleated skirt, plaid skirt, pleated skirt, brown footwear

  • Tracen school uniform :

    • tracen school uniform, purple shirt, pleated skirt, puffy short sleeves, white skirt, puffy sleeves, summer uniform, frilled skirt, sailor collar, sailor shirt, miniskirt, frills, white thighhighs

Use animefull based model

• Settings

  • Stable Diffusion WebUI by Automatic1111

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

  • about 20 steps, CFG scale 3.5~6.5 (monk costume needs lower)

  • kl-f8-anime2.vae

  • CLIP skip = 1

  • Use Hires.fix to get higher quality image

    • Up-scaler : Latent series

    • Denoising strength 0.50~0.65

LoRA training info

• trained on sd-scripts by kohya_ss. Thank you a lot!

• based on Animefull-pruned

• 119 images of el condor pasa

  • 4x low quality 38 images

  • 8x medium quality 34 images

  • 16x high quality 48 images

  • => 1 epoch = 1192 images

• replaced character feature tags with el condor pasa \(umamusume\)

  • horse girl, horse tail, brown hair, blue eyes, etc.

• resolution : 768x768

• rank=alpha=16

• conv_rank=conv_alpha=16

• batch size = 4

• other training settings is included in training_data.zip

This LoRA may be compatible with animefull-based models (ex. AbyssOrangeMix Series...)

This LoRA can generate nsfw images of el condor pasa.

All uploaded images are generated by AOM3A1

Thank you KohakuBlueleaf for introducing LoCon.

And, special thanks to my best friend. DH.