For this week's meditation I felt inspired by the example of N64/NES noise sounds brought up in class. I wanted to create my own interactive noise instrument based on browser interaction sources of entropy. I was already somewhat familiar with white/pink/brown noise in Tone.js, but this gave me the chance to dive a little deeper into the audio noise algorithm and use Web Audio APIs directly.
Tone.js' source code references an algorithm by Zach Denton, which he explains in this blog post. At first I created some buffer sources with a plan to loop through them to play back noise... but I quickly realized this wouldn't work for an interactive instrument which I wanted to react to mouse movement on a web page. So I switched to the more resource-intensive
ScriptProcessorNode approach with a buffer size of 4096 samples (so it runs about 10 times per second with a 44.1kHz sample rate). I created a custom random function which utilizes mouse X/Y position and the current system time in milliseconds.
Here's a short video of me playing around with the instrument:
You'll notice that it quickly gets into predictable cycles if you don't move the mouse. Mouse movement entropy is a big part of this system, so it relies on a "performer" actually moving the mouse to produce a pseudo random sonic aesthetic.