Generating Videos With Scene Dynamics

I'm going to put this link from MIT on Generating Videos with Scene Dynamics right here and hope that others can read and decipher it a bit better than I can. Here's the intro conceit:

We capitalize on large amounts of unlabeled video in order to learn a model of scene dynamics for both video recognition tasks (e.g. action classification) and video generation tasks (e.g. future prediction). We propose a generative adversarial network for video with a spatio-temporal convolutional architecture that untangles the scene's foreground from the background.

From looking at it, these brainiacs are taking a bunch of random videos and creating new scenes based on the dynamic attributes they see within. They are able to generate a Beach, a Train Station and some terrifying babies. I don't have a clear understanding of exactly what is happening but I'm intrigued regardless.