anmnr
※重要 当方の環境での検証の結果、EasyNegativeとの相性が非常に良いことがわかりました。一緒に使用してください。アニメ塗りになるLoRAです。アニメ塗りと言われているモデルよりも本来の意味でアニメ塗りになります。通常、綺麗なアニメ塗りを呼び出すには複雑なプロンプトと当たりが出るまでガシャを回し続ける根気が必要ですが、このLoRAではそれらは不要です。また、教師画像にロゴやテキストが存在しないため、それらが出現しません。0.5-0.8ぐらいで使用するのが良いかと思います。トリガーワードは「anmnr」ですが、あってもなくてもアニメ塗りにはなります。どちらが良いかはよくわからないです。
⋆v1.0 for AOM3A1_orangemixs : AOM3A1_orangemixsで作ったもの ⋆v1.0 for anything-v4.0 : anything-v4.0で作ったもの ⋆v1.0 for Grapefruit v4.1 : Grapefruit v4.1で作ったもの 2023/03/23追加 ⋆v1.0 for Corneo's 7th Heaven Mix : Corneo's 7th Heaven Mix ⋆v2で作ったもの 2023/03/23追加BFCSimplifier test : Better Flat Color Simplifier、"flat color"から"monochrome"や"flat chest"等ノイズを除いたもので、ディテールを削減したい時にごく僅かに混ぜて使います。2023/03/27追加基本的には同じモデル、近いモデルを使ったほうが良い結果になると思いますが、全く違うモデルでも機能します。組み合わせ次第では全く違う絵になります。
※Important: Based on testing in my environment, it has been confirmed that using EasyNegative in combination significantly improves performance. Please use it together.
A LoRA specialized in anime-style coloring. Unlike other models referred to as "anime-style," I stay true to the original meaning of the term.Usually, achieving a beautiful and kawaii anime-style coloring requires a complex prompt and a patient approach of repeatedly spinning the gacha until getting the desired result, but with this LoRA, those are unnecessary.Additionally, logos or text are not included in the teacher images, so they will not appear.
I recommend using me at a value between about 0.5-0.8.
Trigger word is "anmnr," but it's not necessary to use it for me to perform anime-style coloring.
I'm not sure which method is better.v1.0 for AOM3A1_orangemixs was created using AOM3A1_orangemixs.v1.0 for anything-v4.0 was created using anything-v4.0.v1.0 for Grapefruit v4.1 was created using Grapefruit v4.1. Added on March 23, 2023.v1.0 for Corneo's 7th Heaven Mix was created using Corneo's 7th Heaven Mix v2. Added on March 23, 2023.
BFCSimplifier test: Better Flat Color Simplifier.This is LoRA created by removing noise such as "monochrome" or "flat chest" from "flat color". It is used when you want to reduce details. The strength should be very small, such as 0.1. Added on March 27, 2023.While using the same or similar models typically yields better results, different models can also work. Depending on the combination used, the resulting image can be completely different.
★https://civitai.com/models/22977/anime-style-lora-with-better-flat-color★
