User LoRA

XL
Karakter

Roux Louka pilot suit (VXP_Illustrious low cost ONLY, read description) (Gundam, ルー・ルカ パイロットスーツ, ガンダム)

#Gundam

#Roux Louka

#Mobile Suit

Roux Louka pilot suit (VXP_Illustrious low cost ONLY, read description) (Gundam, ルー・ルカ パイロットスーツ, ガンダム) - AI Model cover image
Yüklenme Tarihi
13 Mar 2025 05:17
Kullanım
3.29k
Kazanılan
200,000
Yorumlar
Mükemmel  (3)
Tetikleyici Kelimeler

rouxpilotsuit, rouxhelmet

İzinler
  • Allow image generation & sharing
  • Kullanıcıların modelinizi indirmesine izin verin
  • Ticari kullanımlar

Açıklama

Trained on Illustrious. This LoRA is great for creating Roux's distinctive pilot suit, but not so great for generating Roux's face. It only works well on VXP_Illustrious (low cost version) (https://pixai.art/model/1819999994917251788), especially when combined with the LoRA Roux Louka - Mobile Suit Gundam ZZ (https://pixai.art/model/1840692351539831828).

The emblem will usually be drawn correctly at weights of 1.2-1.6, and her hairstyle and face comes out best with the other Roux LoRA at weight 0.7. Give up on the earring, it won't come out properly.

Used with other models, the pilot suit comes out great but Roux herself is unrecognizable even with other Roux LoRAs. After extensive testing attempting to make other versions of this LoRA with various combinations of training data on various other models, this LoRA and this version is as good as it gets. Basically, it just isn't possible to correctly render the emblem without an extremely high weight. On other models, no matter how the LoRA is trained or with what type of training data used, the weight necessary to properly render the emblem ruins the image on anything other than VXP_Illustrious low cost, which for whatever reason can handle absurd weights.

このLoRAはルーの特徴的なパイロットスーツをしっかり再現できますが、顔の再現度はあまり高くありません。特にVXP_Illustrious (low cost version) (https://pixai.art/model/1819999994917251788) での使用がおすすめで、Roux Louka - Mobile Suit Gundam ZZ (https://pixai.art/model/1840692351539831828) のLoRAと組み合わせるとより良い結果が出ます。

エンブレムは1.2~1.6のweightで正しく描画されることが多く、髪型や顔はもう一つのルーLoRAを0.7で使用すると安定します。ピアスは諦めてください、ちゃんと出ません。

さまざまなモデルで、色々な組み合わせの学習データを使って他バージョンのLoRAを作ろうと徹底的にテストしましたが、このLoRAとこのバージョンが限界です。 正直なところ、エンブレムを正しく描くには非常に高いweightが必要で、それを避ける方法はありません。 他のモデルでは、どんな学習方法やデータを使っても、エンブレムをちゃんと描くために必要なweightが高すぎて、VXP_Illustrious low cost 以外では絵が崩れてしまいます。(なぜかこのモデルだけは、とんでもないweightに耐えられるんです。)

Yorumlar