PixAI・・・基本素材数58枚。V1.10・・・58枚を回転で枚数増やして171枚。V2.10・・・58枚にV1.10の出力画像追加で99枚(回転による枚数の水増しなし)<s>V3.10・・・V2.10の99枚を回転で枚数増やして297枚←過学習っぽくなったので没。削除しました。</s>300枚以上になると過学習気味なって絵柄がくどくなるようです。CheckPointは、Illustrious対応の「VXP」を使ってください。------------------------------------------PixAI: Base dataset of 58 images.V1.10: Expanded the dataset to 171 images by augmenting the 58 images through rotations.V2.10: Combined the original 58 images with outputs from V1.10, resulting in a total of 99 images (no artificial inflation of image count through rotations).<s>V3.10: Expanded the 99 images from V2.10 to 297 images through rotations, but this version was discarded as it appeared to suffer from overfitting.</s>It seems that using over 300 images tends to lead to overfitting, resulting in overly stylized outputs.Please use the "VXP" compatible with Illustrious for CheckPoint.

Nasse from Platinum End|ナッセ from プラチナエンド

This LoRA is a version trained using PixA.artI.

Files with SHA256: 2c1976cd6906856e9b1f6844d99a1667b88039be4e4415d28b19b8f9f4ed08c1