santa hat
deerdeer nosedeer glow
Sign In

Air Groove (umamusume)

435
4.6k
24
Verified:
SafeTensor
Type
LoRA
Stats
4,572
Reviews
Published
Feb 20, 2023
Base Model
SD 1.5
Training
Steps: 1,120
Epochs: 8
Trigger Words
air groove \(umamusume\)
Hash
AutoV2
53378F37D3
default creator card background decoration
Donator Badge
mht's Avatar
mht

LoRA of Air Groove (umamusume). Put LoRA file in your stable-diffusion-webui/models/lora folder and write LoRA notation in prompt to applying LoRA. <lora:your_loha_name:weight>

All of my works is free. If you'd like to support me, feel free to buy me a coffee. 👍

Recommend options

  • LoRA weight 0.65~0.8

  • Trigger words

    • base

      • air groove \(umamusume\)

    • Racing competition clothes :

      • blue skirt, thigh strap, choker, clothing cutout, cowboy shot, cleavage, black shirt, shoulder cutout, o-ring, belt, wide sleeves, black thighhighs, pleated skirt

    • Wedding costume :

      • purple flower, wedding dress, white pants, cleavage, bare shoulders, purple rose, crossed arms, official alternate costume, parted lips, collarbone, shirt, parted bangs, hair flower, see-through

    • Casual clothes :

      • blue jacket, black shirt, brown pants, neckless, brown belt

  • 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

  • 88 images of air groove

    • 4x low quality 22 images

    • 8x medium quality 31 images

    • 16x high quality 35 images

    • => 1 epoch = 560 images

  • replaced character feature tags with air groove \(umamusume\)

    • horse girl, horse tail, brown 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...)

All uploaded images are generated by AOM3A1