Sign In

Seiun Sky (umamusume)

581
6.0k
18
Verified:
SafeTensor
Type
LoRA
Stats
5,954
Reviews
Published
Feb 21, 2023
Base Model
SD 1.5
Training
Steps: 2,672
Epochs: 8
Training Images
Download
Hash
AutoV2
0C320F2770
default creator card background decoration
Donator Badge
mht's Avatar
mht

LoRA of Seiun Sky (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 0.55~0.7

  • Trigger words

    • base

      • seiun sky \(umamusume\)

    • Racing competition clothes :

      • white shirt, short shorts, green shorts, jewelry, choker, bracelet, thigh strap, medium skirt

      • (optional) salute. her signature pose

    • Dance party clothes:

      • white gloves, official alternate costume, white cropped jacket, long sleeves, black socks, kneehighs, white shirt, shoes, vest, white footwear, shorts, buttons, pants, collared shirt

    • Casual clothes :

      • suspenders pants, overalls jumpsuits, black shirt, short sleeves, white footwear, sneakers, one-side off shoulder

      • (optional) fishing

  • 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

  • 144 images of seiun sky

    • 4x low quality 50 images

    • 8x medium quality 46 images

    • 16x high quality 48 images

    • => 1 epoch = 1336 images

  • replaced character feature tags with seiun sky \(umamusume\)

    • horse girl, horse tail, short 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