Sign In

Kase Daiki Style - Pony XL

180
1.2k
2.2k
44
Verified:
SafeTensor
Type
LoRA
Stats
1,242
2,211
222
Reviews
Published
Mar 23, 2024
Base Model
Pony
Training
Steps: 1,728
Epochs: 16
Hash
AutoV2
8AB7FFE489
default creator card background decoration
Legendary Landscape Contest Participant
Cosmos's Avatar
Cosmos

This LoRA aims to replicate the style of artist Kase Daiki.

  • If you like this LoRA, please consider directly supporting the artist (don't forget that without the artist, this LoRA wouldn't exist).

  • Please don't pretend to be or publish in the name of the artist.

Please share you creations! I'm curious to see what you can achieve with my models!

Recommended parameters (Automatic1111 Stable Diffusion WebUI)

Base model:

  • Pony Diffusion XL v6: strongest style expression, best details, sometimes not perfect at anatomy

  • js2Prony: best for anatomy (hands, feet, not deformed body), less details

  • AutismMix: good compositions, may struggle a bit with feet and nylon.

CFG Scale: 6-7

Sampler: Euler a

Steps: >30 (35 - 40 is fine)

LoRA Strength: 0.5 - 1.0

Inpainting is your friend to fix faces and details, but I advise you to use with 20-30 steps.

To generate feet: the model js2Prony seems to work better. Set the LoRA strength to 0.6 - 0.7, then you can use inpainting to fix bad toes.

Recommended parameters (ComfyUI)

Workflow link: https://drive.google.com/file/d/1Wa4A4QMzi3VelPZIL01nvBsYwpmF-0AV/view?usp=sharing

Raw generation:

CFG Scale: 6

Sampler: Euler a

Steps: 30

LoRA Strength: 0.8 - 1.0

Hires Fix:

CFG Scale: 5 - 6

Sampler: Euler a

Steps: 30 - 40

LoRA Strength: 0.8 - 1.0

You'll have to download the custom node Ultimate SD Upscale if you want to use it (https://github.com/ssitu/ComfyUI_UltimateSDUpscale). The workflow comes with two options: basic upscale (works very good) and Ultimate SD Upscale. Choose the one that suits you best.

This model comes with no warranty. Do not use this model for inappropriate purposes.