>>936005705
well, that's a pretty good render, so you should be able to accomplish it. if your WAN setup is up to date, you should be using the self-forcing model, which will allow you to quickly test prompts and see what works and what doesn't.
>"a female horse's rear sways side to side towards camera, her utters move with her motion", perhaps even try a dance lora to see if the motion even somehow translates from the prompt. it's a long shot and will take a good many attempts to get the right motion out of the vid.