This is yet another LoRA for Wan 2.2 image to video. Keyword: wind_skirtNote: though I tried to train wind to do the job model has strong tendency to help with woman's hand. If you do not want that effect stimulate with positive and negative prompt accordingly. Also it depends on starting pose especially woman's body and hand position. And try different seeds. If you not aware you can attach decoder VAE and video combine to high noise sampler "denoised" output and run only high part (press play button on video combiner). Video will not be perfect but movement will be clearly visible, low noise only adds details. So you will basically save half time if you run only high side with different seed/prompt to get desired movement first. Then run low noise to get proper video without changing seed or prompt.

Files with SHA256: 916420cbd60d3f687f7e056775521a548ff1888f2ba18c11a38efcf61f7be2d8