This lora was trained on the I2V model using 25 high quality videos from different angles (behind, side, front and POV), for all your (consensual) ???????? needs. Should be flexible enough that as long as a ???? is in frame, the lora will work with both hands and paddles.
Trained using diffusion-pipe for 3200 steps, 128 epochs. Learning rate of 3e-5.
Try the following prompts to control the action:
A {man|woman} {appears from off screen and|} spanks her ???????? {multiple times|once} with {their hand|a rectangular paddle}, {using big and fast arm movements|gently}.
Workflow and full prompts are attached to the videos.
Based on other Wan loras I've trained it should still work with T2V, but untested.


