Some of my recent research has been using various neural net-based techniques to generate images and videos, most notably Pix2Pix. These are from various projects and training data. Future video and image projects will likely use these techniques as one element.
Trained on continuous perlin noise predicting frames of video taken out a train window.
Various images converted from a dataset of traditional Chinese landscape paintings, mostly from the Song and Yuan dynasties. I particularly love the way the network has understood writing and stamps in a few of them.
Some continuous blue fire.