This LoRA is trained on a single fixed composition and works best at 1024×1024 resolution, matching the training data.

Please follow the structure shown in the sample images — using other layouts or resolutions may reduce output quality or distort the pose.