SDXL, a beautiful girl in light colors and anime style.
This image used WebUI (A1111).
This model is based on 7thAnimeXLB and has been adjusted to produce a transparent, anime-style image.
It specializes in illustrations with soft impression with light colors.
*This model may also be able to generate sexy images, but is not considered as a primary application.
tl;dr
I've written the settings at the end of this page for your reference if it doesn't work.
↓ ↓ ↓ ↓ ↓ ↓ ↓
Model used
Features
The animation may be a bit light since LCM LoRA is merged.
It may be easy to produce “鬼(oni)” horns. Demonic-like horns are also sometimes produced, but often properly.
The background may be westernized in many cases. Please adjust this area at the prompt.
About the future
We would like to move forward with training models to be more faithful to the prompts and to eliminate training data bias in the merged model.
Training the model will require a reasonable amount of time and electricity (or cloud service fees).
We would appreciate your support if you would be willing to help.
//////////////////
SDXLで、淡い色彩でアニメ調の美少女を。
このモデルは7thAnimeXLBをベースに、透明感のあるアニメ調の画像を生成するために調整したモデルです。
淡い色使いで柔らかな印象のイラストを得意とします。
*このモデルは、セクシーな画像も生成可能かもしれませんが、主な用途として考慮してはいません。
使用モデル
特徴
LCM LoRAをマージしているのでちょっと軽いかもしれません。
ちゃんと "鬼" っぽい角が出やすいかもしれません。悪魔的な角もたまに出るけどちゃんとしてる事が多いです。
背景は洋風っぽくなることが多いかも。このあたりはプロンプトで調整してください。
今後について
よりプロンプトに忠実で、マージモデルの学習データの偏りを排除するためにも訓練モデルの作成を進めていきたいと思っています。
モデルの訓練には相応の時間と電気代（もしくはクラウドサービスの利用料）が必要です。
よろしければ支援をお願いします。
setting:
The settings differ depending on the tool, such as so-called webUI and ComfyUI.
I think it would be good if you could set them within the following range.
(Clip skip of XL models can no longer be changed in the latest WebUI, so you may not get the picture you intended. ...ComfyUI is good)
Step: 9 - 24
CGF: 5.0 - 9.0
Clip: 2 (-2)
Description
This model is an alpha version without fine tuning.
Our goal is to develop a model that is as independent of specific character traits as possible, which we believe are currently used in the merge model.