Rules Instead of Notes
Instead of writing every note, you set the rules for picking them. The machine chooses pitches and timing, so the result is different every time but sounds like you designed it.
Stochastic Grains
A grain is a tiny snippet of sound, about 50ms long. By itself it is barely more than a click. But scatter hundreds of them across time and pitch and you get a texture that no human could play note by note.
The trick is probability. You pick a set of allowed pitches, then the machine draws from that set using a rule you choose. "All Equal" means every pitch is equally likely. "Favor Root" weights the tonic so it shows up more often, giving the cloud a tonal anchor. "Favor High" pushes the distribution upward for brighter, more sparkling textures.
Density controls how many grains fire per second, and spread controls how randomly they land in time. At 0% spread the grains are metronomic. At 100% they scatter like rain on a window.
Try it -- press Play and watch grains appear on the scatter plot. Change the pitch rule and listen to how the character shifts. Crank density to 50 for a thick cloud, or drop it to 2 for isolated pings.
Why This Matters
Iannis Xenakis pioneered this approach in the 1950s, building orchestral textures from probability distributions rather than melodies. Today the same idea powers ambient music, generative installations, and game soundtracks that never repeat. You are not composing notes. You are composing the rules that generate notes.
The scatter plot makes the statistics visible. When "All Equal" is selected, the dots spread evenly across all pitches. Switch to "Favor Root" and you see them cluster near the bottom. That clustering is what your ear hears as a tonal center emerging from randomness.