That is absolutely fantastic. I'd like to see the algorithm run on different (higher quality) musicians though to see the results. Is this really an artifact of Nickelback's suck, or is the algorithm just so efficient that it maintains the musicality regardless?
As a hilarious aside, I listened to the remix before the original, and some of the awkward bits I assumed were artifacts of the algorithm turn out to be in the song itself.
Many years ago ('70s I'm guessing) I took a Bach piece and sliced it up via 1/4" tape. Basically randomizing snippets, each about a 2-3 seconds long.
It more or less sounded Bach-ish, but rambling. A few people heard parts of it without knowing what it was and seemed to enjoy it, though I wonder if they had stuck with it they would have realized it didn't have any movement or development. Or reliable meter, for that matter.
Reminds me a bit of Oval's 94 Diskont album, where they scratched the hell out of a cd, and then sampled the bits that would play into strange-but-compelling songs like 'Do While':
It would probably be fairly simple to do this to most Western tonal music (at least now that someone else has done the hard part and written a tool that automates the process!), as the biggest factor I see in keeping the sound coherent is the timbre and texture of the sound, not necessarily the actual music.
Most pop/rock/etc music has (at most) a single key change, which means that 75-80% of the song uses the same 8 pitches over and over again. Add to that the rather homogeneous sound of modern pop/rock/etc instruments (not to mention the limited range of vocal styles usually applied within a single song) and I'd guess this would work with most "popular" music written over the past 50 or so years.
There would certainly be exceptions (Bjork, for example, probably has enough textural vocal changes within a single song to make this much more difficult), but by and large, I'm guessing the output of this experiment is more a commentary on modern pop music theory than it is the specific stylings of Nickleback.
Fun example. And the idea of a Fractal track is one of those: how-haven't-I-thought-of-it-before things.
I don't know how a true fractal song would sound. The author of the post uses the original song to smooth the final result to make it listenable. I guess a true fractally generated track could/would have infinitely high pitches that could make it unpleasant for a human's ear.
It'd be interesting to see what would happen if you applied this to all of Nickelbacks work and limited it to about 3 minutes 30 seconds and see if what pops out sounds like one of their songs.
