Music Machinery

Ha, thanks for the shout out! It really is pretty wrong, but I was surprised that it came out as semi-listenable in parts as it did.

This came on the tail end of a lot of fucking around with afromb the last week or so where I found that it’s pretty hard to get anything coherent out of a mix=1 pure-b-track composition; multiplexing it like this helped some in terms of making it more texturally interesting than a single afromb run, but that doesn’t deal with the problem of the output tracks all being confined to the same sometimes-hurky-sounding beat analysis of the a track — the new rhythmic (or often somewhat arhythmic-seeming) signature of the analyzed track gets manifested by every run, so you don’t get more complicated sub-beat stuff that varies between each run.

NIN as source material works better than a lot of stuff in that respect it seems like, I’m guessing because the songs have such aggressive uptempo sub-beats that give the beat analysis routine a really granual picture of the original song. And so the output is more explicitly machine-gun rhythmic, and you get fewer weird stumbles and dead-air gaps from short sharp attack bits from the b track being paired to longer smoother bits form the a track.

Anyway, I’m loving this stuff. Have a bunch of other ideas I want to play with yet; I need to get up to speed on Python and the API so that I can start fiddling more at a code level than I have been so far, start turning my What If stuff into Why Not &c.