I think the hardest part is getting the reference images and using a model like seedance for the motion. I've tried tons of different prompts with the latest models and never can get images as good as these videos. Whole frame makes sense, pr…