LoRA of Tamamo Cross (umamusume)

NOT COMPATIBLE WITH AOM3A2, AOM3A3 model.

Recommend options

• LoRA weight 0.7~1.0

• Trigger words

  • base

    • tamamo cross \(umamusume\)

  • Racing competition clothes :

    • navel, midriff, crop top, fingerless gloves, blue footwear, belt, open mouth, blue jacket, white pants, white gloves, sneakers, collarbone

    • (optional) : light trail, glowing eyes, electricity, glowing

  • Casual clothes :

    • collarbone, shirt, long sleeves, denim shorts, sneakers

  • Festival costume :

    • ponytail, hip vent, official alternate costume, sleeveless shirt, jewelry, black footwear, clothing cutout, black pants, black shirt, bridal gauntlets, baggy pants

  • Sailor costume :

    • sailor collar, neckerchief, pleated skirt, wide sleeves, black skirt, belt, collarbone, long sleeves, grey thighhighs, blue shirt, loafers

  • Aged down version :

    • kindergarten uniform, school hat, bag, white socks, bandaid on knee, black shorts, blue shirt, long sleeves, aged down, yellow headwear, white footwear, ears through headwear

    • (optional) drooling

Use animefull based model

• Settings

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

  • about 20 steps, CFG scale 3.5~6.5

  • kl-f8-anime2.vae

  • CLIP skip = 1

  • If using highres

    • Latent series

    • Denoising strength 0.50~0.65

LoRA training info

• based on Animefull-pruned

• 163 images of tamamo cross

  • 4x low quality 45 images

  • 8x medium quality 62 images

  • 16x high quality 56 images

  • => 1 epoch = 1572 images

• replaced character feature tags with tamamo cross \(umamusume\)

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

• using aspect ratio bucketing

• resolution : 768x768

• rank=alpha=16

• using dadaptation

• batch size = 4, 8 epochs trained

• clip skip = 1

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

This LoRA can generate nsfw images of tamamo cross.

All uploaded images are generated by AOM3A1