etc

Amy and I are currently residents at the Takt Residency in Berlin. One of the projects I’ve been working on, The Invisible Orchestra, is equal parts social experiment, algorithm, and music. From the website:

The Invisible Orchestra is an electronic music experiment. Using a collection of computer algorithms that rely–to an extent–on randomness, each week (or month?) we generate a collection of “riffs”–6 melodies and 4 drum tracks–for you to create original compositions with. Some riffs are more random than others. Some weeks’ riffs will be more conventional and listenable. Some will be more…well…difficult.

You–the “composers”–will build compositions out of computer-generated “riffs”. Riffs are simply midi-files, which include no information but which notes to play or drums to hit, how loud to play them, and for how long. Composers choose the voice and arrangement for these riffs. Sections might be looped, silenced, effect-laden, cut into pieces, etc.

The songs might be barebones, or they may include original (human-generated) riffs.

From week to week, we might tweak both the algorithms that generate these riffs, as well as the Rules of Composition (must composers use all available tracks? are human compositions okay? outside samples? vocals?)

Here are a handful of songs and early experiments, mostly by me, but including a growing list of other musicians and artists (shout out to Keith McDougall, Yuria Okamura, and the mystery composer Death By 7 Digits):

For those interested, I’ll post the (Open Source/Creative Commons) algorithms to Github in the next couple of weeks.

I’ve been working on and off for a while now on a project to create an unending song, compiled on the fly from a selection of 30 second loops of individual instrument tracks: electric guitar, bass, electronic drums, electronic piano/keyboard.

http://songthatneverends.nestsounds.com:8000/ices.m3u

The concept is fairly simple

  • compile a directory of 30 second song parts, separated by instrument
  • combine this algorithmically to create a full song section (so, guitar, bass, drums, maybe piano, maybe two guitars)
  • vary some–but not all–parts, to create a sense of continuity and flow

While most of the really tough technical aspects have been ironed out, it still isn’t super effective, probably because it (1) needs a lot more parts, and (2) the algorithm still needs some fine-tuning in terms of the rules it uses to assemble each successive section.

That being said, you should be able to let this play ad infinitum. You just might want leap out a window after about 20 minutes.

Poke back in for updates.

Back to top