Sign In

Leah illustriousXL (NoobAi) (Airisubaka | EIton) | LoCon

25

81

6.7k

1

Verified:

SafeTensor

Type

LyCORIS

Stats

81

6.7k

5.7k

Reviews

Published

Nov 13, 2024

Base Model

NoobAI

Training

Steps: 1,560
Epochs: 10

Usage Tips

Clip Skip: 2
Strength: 1

Trigger Words

leai2airii, orange eyes, brown bandaid on cheek, short hair, purple hair, high ponytail, large breasts
two-tone headwear, grey headwear, baseball cap, hat ornament
Short hair, high ponytail
Short hair, low ponytail
Long hair

Hash

AutoV2
E0D1D137DA
default creator card background decoration
Chameleon Badge
Manityro's Avatar

Manityro

The OC tomboy Leah from "Airisubaka". Works best at 1.0 weight. Trained on AAAAAA-Finetune (NoobAi-1.0). Request form

Warning: As this character is trained on a NoobAi based model it may not work as intended on illustriousXL-0.1 based models. I recommend using a NoobAi derived model, there are some in the suggested resources.

Always needed:
"leai2airii, orange eyes, brown bandaid on cheek, short hair, purple hair, high ponytail, large breasts"

  • Heavily suggested: "two-tone headwear, grey headwear, baseball cap, hat ornament"

  • A ponytail is her main style but it can be changed combo ideas below:

    • "Short hair, high ponytail"

    • "Short hair, low ponytail"

    • "Long hair"

Outfits/notes:

  • Check samples above for outfit/prompt ideas.

  • The cheek bandaid can sometimes gen as if there were 2 but compared to before it very stable and a reroll tends to fix.

  • Flexible and nice data variety and illustrious gives really good flexibility.

    • Note I cooked the hat in a bit stronger so may need to negative "baseball cap" if it bleeds over other hats.

At first I thought illustrious was not great since looked ugly but then NoobAi released and I made myself a checkpoint to enjoy and oh my it is SUCH a breath of fresh air after Pony. I been having a lot of fun so I hope everyone else can as well as I start to port over various characters. I hope you all enjoy.

Feedback and reviews are always appreciated.

Training data is in the Lora metadata.

Trained on 86 images.