Donut Age: America's Donut Magazine

Beyond iTunes

Not that my love of iTunes and the ITMS is waning, but recently, I've been experimenting with a couple other online music services. Pandora and eMusic are quite different from each other, and for my purposes, both are complements to rather than replacements for iTunes. Each is interesting in its own way.

Free as in freedom

Of the two, competes more directly with Apple's music store. It's a subscription service, but unlike, for example, Napster, the subscription fee entitles you to a certain number of free downloads (basic plan is $9.99/month for 40 downloads). In addition, unlike both ITMS and subscription services, the downloads are unprotected MP3s, so once the music gets to your computer, it's yours to do with as you will. Finally, eMusic claims to specialize in independent music. The latter may just be making the best of the fact that major labels would never consent to releasing DRM-less music, but preliminary browsing has certainly turned up some desirable albums that still haven't made it onto ITMS's shelves. Even where there is a choice between ITMS and eMusic, the cost (effectively 25¢/song) and the unprotected file status make eMusic an attractive alternative. The sound quality also seems pretty good: the files use variable bit rate encoding with averages (so far) of 150-230kbps.
eMusic also incorporates a number of 'social' features—keeping track of your 'friends' and 'neighbors', offering suggestions of people with similar tastes— but I have not explored these yet. Perhaps the handiest feature I've found is 'Save for Later', which lets you set aside albums for future reference. I like this better than the ITMS approach (you can 'bookmark' any song by dragging it to a playlist, but there's no way to reference an album per se), although because of eMusic's pricing model, I'd actually be more likely to want to mark individual or selected songs from an album there than ITMS (where the price-break for buying a complete album often negates any savings you might get from excluding particular songs).
In summary, though, I'll be sticking with eMusic for the immediate future at least. I kind of like the pressure to acquire more music that the subscription models applies, and as long as I can keep coming up with 40 songs to bring home each month, it's a worthwhile deal.

Find Music You Love

Pandora is a quite different case. It does not sell music but rather bills itself itself as a music discovery service. It's an offshoot of the Music Genome Project, which has spent the last several years attempting to catalog the fundamental formal and aesthetic characteristic that make up the 'genes' of music:
On January 6, 2000 a group of musicians and music-loving technologists came together with the idea of creating the most comprehensive analysis of music ever.
Together we set out to capture the essence of music at the most fundamental level. We ended up assembling literally hundreds of musical attributes or "genes" into a very large Music Genome. Taken together these genes capture the unique and magical musical identity of a song - everything from melody, harmony and rhythm, to instrumentation, orchestration, arrangement, lyrics, and of course the rich world of singing and vocal harmony. It's not about what a band looks like, or what genre they supposedly belong to, or about who buys their records - it's about what each individual song sounds like.
Over the past 5 years, we've carefully listened to the songs of over 10,000 different artists - ranging from popular to obscure - and analyzed the musical qualities of each song one attribute at a time. This work continues each and every day as we endeavor to include all the great new stuff coming out of studios, clubs and garages around the world.
Pandora makes use of this body of information by identifying songs with similar 'genetic' features. When you sign up, you are asked for the name of a song or artist you like. Pandora then creates an online 'station' that streams music with that song or artist's features. As you listen, you can give the system more information by giving songs a thumbs-up or thumbs-down as they play or just adding names of favorite artists to your profile.
So far, Pandora has shown some aptitude for finding music I like (it just picked a nice Blake Babies song for me without having been told that I like the band), but it has also produced some real howlers (like when it tried to offer me Shaun Cassidy). It sometimes seems like their matching algorithm will get kind of hung up on a particular feature (as when I got a string of songs notable for their use of 'prominent organ'), but it is hard to tell for sure what's going on behind the scenes.
I have some mixed feelings about this approach. On the one hand, it seems to hold more promise as an actual discovery mechanism than the "other customers also bought this" recommendation engines used by Amazon and the like. However, there also seems to be some thing mechanical about reducing aesthetic appreciation to a checklist of formal features. Out of curiosity, I have been keeping a list of the 'genetic' features that Pandora says it has used to select my music. The most frequent ones mentioned are:
  1. Mild rhythmic syncopation (84, I'm pretty sure it has showed up on every song I checked)
  2. Major key tonality (53)
  3. Mixed electric & acoustic instrumentation (49)
  4. Extensive vamping (37)
  5. Electric rock instrumentation (32)
  6. Subtle use of vocal harmony (30)
  7. Repetitive melodic phrasing (23)
I'm not sure how meaningful this information is. First, I'm certain Pandora does not list all the features that contributed to a particular selection (it typically lists three to five and sometimes will throw in a phrase like "and many other features"). The most frequent features strike me as being typical of entire genres rather than identifying particular styles or approaches. Some interesting traits have cropped up in the lower-frequency features—"groove based compositions," "gravelly male vocalist," "political satire lyrics"—but at one instance each, I'm not sure these can be said to be very meaningful either. It would probably be pretty fascinating to get under the hood of Pandora to find out the full range of features it identifies and how it really weights them in the matching process, but I assume that such things are closely guarded secrets. In any case, it's interesting enough to serve as an alternative to iTunes random play, which provides my usual soundtrack.