style by hara_(harayutaka)

Yodayo / (sd 1.5) | Seaart.ai

Versions for sd 1.5, Animagine and Pony, contains several for each, pick whichever you think you need. I recommend Animagine. Thousandths in ver. are just model differences, the dataset is identical

For all versions:

Minimum recommended steps: 25-28

Trigger word: shr

 
All images have a prompt you can look at and choose what you are interested in!

v0.91 PDXL

Trained on Pony diffusion v6.

Recommended weight: 1-1.5 (1.35)

Recommend word: (hara \(harayutaka\):0.45-0.7) ~ (0.5)
When using this word, it is recommended to reduce the weight of lora.
It is not recommended to use the word without lowering the weight.

Recommended word for harayutaka original style: anime screencap

Sweet spot epoc: 12

v0.911 SDXL

Trained on Animagine XL v3.1

Recommended weight: 0.8-1 (0.9)

Recommend word: hara \(harayutaka\)
When using this word, it is recommended to reduce the weight of lora

Sweet spot epoc: 09-11 (11)

Differences: more original style, perhaps because the model knows the original style and has it inside it.

v0.932 sd 1.5

Trained on aamAnyloraAnimeMix

Recommended weight: 0.75-1.1 (0.9)

Sweet spot epoc: 13-16

Differences: more flexibility in everything because of the large set of checkpoints and lora sd 1.5.