Mapping Music

I’m a huge music nerd.

I’ve had an image in my head for a while. I sent a note to my friend Alex saying something about how I’d like to walk along a landscape of music wherein:

  • Things that people listen to together are close together
  • Things that get listened to more are higher

There’s an ever growing mountain ridge of top 40. Lower down: Geometric black metal gardens of minimal techno. The twee valley of Rockabilly.

Wander around and find a record label. Find musicians who went to high school together. Find cities, countries, scenes. Grooves of a certain BPM.

OK that’s pretty hard to do, so I figured this weekend I’d bite off a much simpler but still pretty neat project in that direction—hack something together on top of the excellent Echonest API that does a graph visualization of relatedness and their notion of “hotttnesss”, which tracks popularity. Sorta.

Thankfully, they have a nice python client library and there are some sweet graph visualization tools out there these days.

I decided to use color and size to represent hotttnesss and some reasonable layout algorithm to try and shake out nearness. Roughly.

With a few hours of poking around, I give you a map of bands that sound like Boards of Canada:

I’ve heard of all the big red circles: Caribou, Tycho, M83, Bonobo (all great!). The fun ones are the small blue discs; bands like SeeFeel or Gescom. You’ve all got homework!

For good measure, I tried out a slightly less popular band, Holy Other. So here’s what I’ll claim is the first map of Witch House:

There are a few big reds on the edges, but mostly a morass of low energy balls with weird unicode characters. Huzzah!

Tiny bit of echonest code is here.