Okay, I have a theory about a way that a video could work with dynamic depth, that said it would take a lot of work and Nintendo would actually make a video player to support it.
The video would need to be filmed in layers, let's visualize three things.
============================
A back wall, this is the farthest back part of the picture, let's say it has a picture on it.
++++++++++++++++++++++++++++
This is the mid point, let's say there is a person sitting here.
--------------------------------------------------
This is the front part, let's say there is a desk here with stuff all over it.
Now to make this work each section needs to be filmed separately, putting the person and desk in front of green screens. Now we have three layers filmed with a camera made for 3D recording. It would have to be encoded in a special way to allow for the layers to still be independent of each other but still appear over each other with transparency added to the mid and front layers. Then when played on the 3DS each layer could be moved together or farther apart depending on the depth level, just like in games.
I expect at some point we'll see videos like this made for newer 3DS games after companies have time to play around with it. I don't foresee Hollywood movies doing this though, at least not unless there is a way to do the same thing for 3D TVs too.