ETHAN MARSHALL COHEN

As part of my work with Rockwell Ventures, I consulted for Bose on the design of a generative audio system for their Sleepbuds II hardware project. 

 

The Sleepbuds were a unique approach to audio hardware, eschewing the traditional functionality of headphones. Their form-factor was designed fit completely within the profile of the pinna (outer ear), such that a user could sleep comfortably on their side without any pressure or discomfort resulting from wearing them while sleeping. This design constraint resulted in a host of knock-on effects, primarily that most of the hardware's miniscule size was simply taken up by batteries that would be able to last overnight. Even then, Bluetooth streaming over long periods of time was untenable, as it demanded too much use of the already very limited battery life.

 

Thus, the unique design solution was to have the hardware itself act as the audio player; instead of the earbuds streaming signal continuously from another source, like a phone or computer, the buds themselves would store the audio content natively and act as playback hardware. Here too, the form factor caused extreme limitations: there was simply not enough storage space on the device to host much more than a few minutes of audio with acceptable quality, let alone a whole night's worth. As of my consultation, the design team hoped to limit the audio to ~120s.

 

Thus, my task was to design a generative music system that, with audio files not exceeding a total of 120s, could play back unique and effective music over much longer periods of time - as much as 8 hours! Here, there were not only the aforementioned hardware constraints at play, but also the parameters of effective music for rest our team had come to understand over years of finding and understanding the clinical evidence. One of the primary findings of this research was the notion of 'listening fatigue' - that is, the stimulation of awareness and attention in a listener when a sound source becomes repetitive and, ultimately, boring. Thus, it was clear from the onset that simply creating a 120s piece that only looped over and over throughout the night would not suffice (and in fact would function antagonistically to our goals). 

 

We knew that the subtle interplay of repetition and difference - familiarity and variation - was the key to a successful piece of music or sound for relaxation and sleep. And, using the framework provided by a groundbreaking clinical study on the formal elements of music that induce physiological and psychological calm, I designed a system that consisted of twelve 10s clips that could seamlessly transition from one another. The sections were divided into 5 positions, each position having one or two variations. The system would thus move through each position, selecting from one of the variations at random, and sometimes even selecting which position to move to next at random, creating an infinite-yet-evidence-based playback sequence out of those 12 audio clips, transforming the constraint of only 120s of total audio into an ever-evolving yet comfortingly-familiar soundscape.