Sign In

[Illustrious] Encore (Wuthering Waves)

127
873
977
35
Updated: Mar 16, 2025
characterwuthering waves
Verified:
SafeTensor
Type
LoRA
Stats
484
966
457
Reviews
Published
Mar 12, 2025
Base Model
Illustrious
Training
Steps: 3,050
Epochs: 25
Usage Tips
Clip Skip: 2
Strength: 1
Trigger Words
encore \(wuthering waves\), cross-shaped-pupils, thick eyebrows, purple eyes, blush, hair ornament, pink hair, ahoge, hair bow, pink bow, low twintails, dress, bag, earrings, short, black sleeves, white sleeves, bowtie, black footwear
encore \(wuthering waves\), cross-shaped-pupils, thick eyebrows, purple eyes, blush, hair ornament, pink hair, ahoge, hair bow, pink bow, low twintails
Hash
AutoV2
7CA2CF40DC
default creator card background decoration
vjumpkung's Avatar
vjumpkung

Encore (Wuthering Waves)

Version 1.1 description

  • LoRA+ with CLIP-L Only no more LoCon

Version 1 description

  • First Lycoris LoCon

  • using Lion Scheduler

  • Pink hair girl again.

Trigger Word

  • default costume (weight 0.9 - 1)

encore \(wuthering waves\), cross-shaped-pupils, thick eyebrows, purple eyes, blush, hair ornament, pink hair, ahoge, hair bow, pink bow, low twintails, dress, bag, earrings, short, black sleeves, white sleeves, bowtie, black footwear
  • any costume (weight 0.8-1) maybe some costume cant change

encore \(wuthering waves\), cross-shaped-pupils, thick eyebrows, purple eyes, blush, hair ornament, pink hair, ahoge, hair bow, pink bow, low twintails

Limitations

  • default costume still inaccurate sometimes

  • checkpoints maybe know this character but extremely inaccurate.

Training Details

LoRA size

  • Dimension 16

Dataset

  • 80 images

parameters

  • LoRA+

  • resolution = 1024

  • batch size = 2

  • Network dim,alpha = 16,16

  • mix/save precision = bf16/bf16

  • optmizer = Lion

  • UNet LR = 1.75e-5

  • TE LR = 1e-5

  • scheduler = cosine_with_restarts

  • l2 loss

  • unet + clip-l

steps

  • total steps = 3050

tools

  • kohya-ss GUI v24.4.0 (custom version)

  • torch 2.5.0 cu124

  • Runpod RTX 4090 24 GB + xformers + gradient_checkpointing

weight

  • UNet average weights : 0.011061001399701292

  • TE1 average weight : 0.009004636706473926

*This LoRA is for studying LoRA training with new technique so do not use for damaging the character (also support Wuthering Waves too).

*REPOST model is not allow if you found this model please report to me.