Given a set of bounding boxes with associated trajectories, our framework enables object and camera motion control in image-to-video generation by leveraging the knowledge present in a pre-trained ...