This workflow uses an image as reference, and prompt traveling to shape the output video more exactly than when only using Animatediff.

This is a work in progress and further development will be done.