One thing that I think could be done better would be to display the BPM Deviation by percentage instead of a hard tempo.
This is especially important when there is ambiguity over what the tempo actually is. Is this song 90 BPM or 180? That question is pretty hard for a computer to tell, and two great musicians might tell you different answers. It's pretty subjective.
Here's an example of two takes of a slow rubato jazz piece. The computer analyzes one using a tempo of 169 BPM and the other at 91. It's the same song, and a very similar tempo. But the one rated faster looks much more erratic because the deviation isn't displayed as a percentage.
This is fascinating. Of course most modern music is generated with click tracks, if for nothing else than it makes it easier to edit in Pro Tools, since each track will be aligned with a beat grid, and you can literally cut and paste entire bars of music.
This has more to do with the modern way of generating music - get four or five takes of the song from each musician, take the best bars and phrases from each one, and put them all together. Maybe Britney had a great take on the second take of the chorus, so you just copy/paste that 5 times and now every time the chorus plays it sounds perfect. Maybe the instrumental solo was perfect the third time, so let's use that one.
This also makes you have great appreciation for those bands that recorded music before the advent of modern multi-track recording software, and were somehow dedicated enough to get that one perfect take where everyone in the band did great and nobody made a mistake.
The guys at echonest seem to slip through the cracks a bit on HN, but they are doing some amazing stuff. I encourage anyone with an interest in music to check out the API.
Those are some really fascinating results. Although now and then you'll see machine-like scores when there's no way. Most interesting is to look at old-timers that boozed it up alot... tempos ALL over!
It looks like Neil Peart uses a click track on Stick it out
or he could just be rock solid.
I wonder how well Tico Torres and Billy Cobham (great drummers and time-keepers) backed tracks fare, especially one from the 80's and early 90's when computer editors were less common.
A quick click through also suggests that Bon Jovi tracks are click-tracked.
It's worth bearing in mind that click tracks have been around since the early 20th century, none of this requires computers, just headphones, though using computers for editing means you get an extra benefit from click-tracks (cut'n'paste with consistent beat) which might push more people towards them.
Metronomes of course, have been around forever. As early as the 70's a lot of bands started syncing to an electronic source. Especially the prog bands, with excellent drummers, who were versatile enough to sync at will to the arpeggiators on analog synths. And even those arpeggiators often had tempo drift due to the analog electronics.
I looked in the archive for this article -- an amazing collection of data by the way -- and found U2 tracks with tempo variations all over the map:
This project has created some amazing data, but I don't agree with all the conclusions at face value. Some of the stuff that is labeled as too machine-like, is well within the discipline of a lot of drummers. Meanwhile modern click tracks can be easily programmed to speed up and slow down throughout a track. In fact, it's done all the time.
EDIT: One other thing, some of the most interesting intentional tempo variations occur in the cycle of measures, phrases, and sections. The data presented here, only measured in seconds, doesn't track that relationship which would be the most musically interesting thing. It would be a good experiment to listen to some of these tracks with a prompt following the timeline in real time.
This chart looks pretty crazy at first, but if you are familiar with the music it makes so much sense. I believe that they (the entire band) knew exactly what was happening with the tempo throughout the entire 15 minute song.
The most interesting results for me are Charlie Watts'. Of the Rolling Stones' explicitly dance-oriented tracks from the 70s, "Hot Stuff" and "Miss You" are 87%, and "Shattered" is 85%. I read this as being way too loose to have had machine help, but way steadier than just about any human drummer.
One thing that I think could be done better would be to display the BPM Deviation by percentage instead of a hard tempo.
This is especially important when there is ambiguity over what the tempo actually is. Is this song 90 BPM or 180? That question is pretty hard for a computer to tell, and two great musicians might tell you different answers. It's pretty subjective.
Here's an example of two takes of a slow rubato jazz piece. The computer analyzes one using a tempo of 169 BPM and the other at 91. It's the same song, and a very similar tempo. But the one rated faster looks much more erratic because the deviation isn't displayed as a percentage.
http://labs.echonest.com/click/?trackId=TRFQMHM123E8585C4A&#...
http://labs.echonest.com/click/?trackId=TRGHGWJ123E858DF63&#...