aleaさんのイラストまとめ

まひろが可愛いすぎと思って泣いているヒナギク
(今日はヒナギクと緒山まひろ役の高野麻里佳さんの誕生日です、お誕生日おめでとうございます。)
https://t.co/0koMrQDpfY
#おにまい
#おにまいお絵描き
#くノ一ツバキの胸の内
#AIイラスト #AIart

エイラとぼっちちゃん誕生日おめでとう
本日の主役も二人ですね
#後藤ひとり生誕祭
#エイラ生誕祭2023 #エイラ生誕祭
#AIイラスト

Notice that the extracted lora does work, e.g.
They are just (in my opinion) not as good as merging with the original model
I hesitate in extracting loras from other models and share them (everyone can do that actually)

1. orange+onimai model-AC
2. orange+ExtratLora(onimai model-AC)
3. orange+trained onimai lora
I believe not having the weight difference of the residual blocks is very detrimental
This is also why I always train entire model and merge them
#stablediffusion #lora

あおいちゃんとちよのおーちゃん、同じ誕生日ですね
https://t.co/I4HsBSGFf4
(working for a stable diffusion model for the entire umamusume, hopefully coming soon
#サクラチヨノオー生誕祭2023 #ウマ娘
#AIイラスト

あおいちゃん誕生日おめでとう
#ヤマノススメ
#雪村あおい生誕祭2023
#AIイラスト

To make this post self-compete, I am posting here again the euphi-ani image that I use as input, and sur ce, I am going to continue to read the novel.

If we take image A as input and try to match the low frequency of image B. I cannot say what happens but sometimes we can get interesting results.