CreateJapaneseAnimeStyle ALL+NSFW
#Male #Boy #Girl #Furry
【BASE】
VAE data is considered better with
This is recommended
Anything / Kl-f8-anime2 / Vae-ft-mse-840000-ema-pruned / Blessed / ClearVAE (fp16/cleaned)
vae-ft-mse-840000-ema-pruned | 840000 | 840k SD1.5 VAE
Furception VAE v1.0, by Project RedRocket
Recommendation.
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 7
Model used for merging
MY LOVERY LORA
-Taniguchi-
https://civitai.com/models/480574/original-japanese-face-anime-style-taniguchi-sdanimatediff
-Jyunzaburou-
https://civitai.com/models/480435/original-japanese-face-anime-style-jyunzaburou-sdanimatediff
These are merged with my drawing style into a Meina＆Yiffy, and I am not affiliated with any other illustration authors.
No profit making of any kind is permitted.
These terms apply to the model of origin
and to the merging of my models
Description
This is a very good excuse.
I started using SD1.5 to run AnimateDiff, but for some reason YiffyMix spits out noise while MeinaMix works well with it, so I created my own merging model for both.
In order to differentiate MeinaMix from YiffyMix, I mixed in Lora, which was trained with my own original patterns.
It took a lot of trial and error to get it close to my own pattern.
This is a checkpoint that I did a lot of work on.
めっちゃ言い訳になりますがw
AnimateDiffを動かすのにSD1.5に手を出したもののMeinaMixは相性いいのにYiffyMixはなぜかノイズ吐くので独自で双方のマージモデルを作成しました。
MeinaMixと差別化を図るために自分の独自の絵柄を学習させたLoraを混ぜました。
自分の絵柄に近づけるのに何回も試行錯誤してめっちゃ大変でした💦
かなり力作のチェックポイントです！