Here's a quick clip with an idea for how the clouds could come toward the viewer. It still looks a bit stiff and artificial. I could also animate individual images within the video. Here I've just used changing the speed, direction, and center of an existing clip.
I imagine the interactivity could be done either by motion sensor or by sensors activated when people step on them. Motion sensors would be great because then movement of arms or more dramatic body motion could create imitative actions in the video.Besides for interactive video, I could also create a sculpture of a tree or clouds which move or come toward the viewer when similar motions are performed.