This is an experiment to train a LoRA for video, turns out for transformation only High Noise needs training for WAN. Wanted to have something that doesn't have the unwanted motion, pose changes and disappearing clothes of existing LoRAs. Sadly clothes still have a tendency to disappear