V2.0

1. Trained over AnyLora Anime Mix

2. Should be more flexible with more models than just Anything V4.5

3. Does not replicate Fumio's style as closely as V1.0 but then again should be more flexible.

4. Both LoRA and dataset are 768x768px

A lora based on "Suou Amane" from Grisaia no Kajitsu and its sequels a lovely girl, if you like this model please consider buying the novels which feature her!

I used a lot of CGs from the novels even some art that's only in the books that come with the Sofmap edition, I hand picked the images and processed them to focus on Fumio's style but some of the anime art is in there as well and some CGs drawn by Akio Watanabe, model works fine from 0.8-1 but the pics are taken with 1.0

Also thanks to @rinrinrinrinr for their Nicorima LoRA they inspired me to do this one, also copying his note here

Training settings?

The most important part of training is getting a good dataset!

Training settings can be found by installing the additional networks extension in auto1111

Extensions -> available -> load from -> additional networks

Restart webui

place lora into stable-diffusion-webui\extensions\sd-webui-additional-networks\models\lora

To view the training settings

Go to additional network tab -> choose model -> training parameters

描述:

First version, trained with images with varied resolutions up to 1024x1024

训练词语: