Take a Trippy Journey in This Machine-Generated Video
Coder Damien Henry created a 56-minute film based on one starting image and a machine learning algorithm. He trained the machine using video shot from the window of a moving train. Then he handed off that first frame and had the algorithm generate what it thought might be a "next" frame. That process repeated serially for the entire film, resulting in this beautiful, abstract train ride:
The soundtrack is Steve Reich's classic "Music for 18 Musicians," and the pairing helps create a mesmerizing atmosphere. Watching the video, you see smears of light and dark eventually form into landscapes (drawing on the algorithm's knowledge of landscapes), but those landscapes are often messy and surreal, looking like blobs in a lava lamp or perhaps a robot's low-fi idea of what a landscape might look like. Because the film includes zero editing, it is purely a product of that first frame and the machine's training. It's beautiful.
In the YouTube description, Henry wrote (in part):
The results are low resolution, blurry, and not realistic most of the time. But it resonates with the feeling I have when I travel in a train. It means that the algorithm learned the patterns needed to create this feeling. Unlike classical computer generated content, these patterns are not chosen or written by a software engineer.
In this video, nobody made explicit that the foreground should move faster than the background: thanks to Machine Learning, the algorithm figured that itself. The algorithm can find patterns that a software engineer may haven’t noticed, and is able to reproduce them in a way that would be difficult or impossible to code.
He also notes that the algorithm learns during the video's creation, which accounts for the increase in realism as the video goes on. He notes that the algorithm's learning system is updated every 20 seconds.
For a bit more from Henry on the project, check out his Twitter feed.