X

Mufin Player organizes songs by sound

The Mufin Player recommends similar songs based on an audio analysis, but the first beta version has a litle trouble identifying song information from iTunes.

Matt Rosoff
Matt Rosoff is an analyst with Directions on Microsoft, where he covers Microsoft's consumer products and corporate news. He's written about the technology industry since 1995, and reviewed the first Rio MP3 player for CNET.com in 1998. He is a member of the CNET Blog Network. Disclosure. You can follow Matt on Twitter @mattrosoff.
Matt Rosoff
2 min read

Earlier on Friday, Mufin launched its music player, which analyzes the songs in your music collection based on their audio content, rather than on human analysis or genre.

Human analysis, naturally, is subjective, and genre labeling is totally arbitrary and unreliable--I particularly hate the meaningless label "Alternative," which can apply to everything from dead-slow acoustic to fast punk; it's more about the hairstyles of the artists than the content of the music.

Once Mufin has analyzed your tunes, it can recommend similar-sounding songs from your collection. It also catalogs songs in its online database and can recommend music from its own sources, if it doesn't find anything similar on your hard drive.

My Mufin Player installation went smoothly, though it took a little more time than I expected, because I had to download and install the open-source SQL Firebird desktop database. But when I began importing my music collection, Mufin could only recognize metadata from files that had been originally downloaded as MP3s.

Anything I ripped or converted through iTunes simply showed up as a file name. I suspect that there is some sort of incompatibility between how iTunes and Mufin use the ID3 metadata format, but who knows? The Mufin help file offers no help on this matter.

Mellow, meet more mellow.

Mufin took quite a while to analyze my collection--as I write this, it's been plugging away for about 20 minutes and is through about 700 of the 3,400 songs on my hard drive--but once it got a reasonably large data set, the recommendations were spot-on.

For instance, I selected "If You Want Me," a downbeat acoustic-guitar-laden track by Marketa Irglova and Glen Hansard from the "Once" soundtrack. Mufin selected Beck's "Sing It Again" as the most similar, followed by Beck's "End of the Day" and "Already Dead." (I'm sure the results would have been more diverse, if the program had had time to analyze a larger portion of my collection.)

For my second test, I had it analyze Cuong Vu's avant-jazz piece "Expressions of a Neurotic Impulse." Mufin's top similarity: a live version of "How I Learned My Lesson" by X, followed by the Butthole Surfers' "Some Dispute Over T-Shirt Sales."

Acid jazz and punk? The songs do actually sound similar--noisy, abrasive, lots of cymbal. This is the kind of unexpected recommendation I was hoping for, based on the sound of the music rather than the mostly artificial "genre" categories that artists and labels pick.

Once they work the ID3 bugs out, Mufin could become a useful piece of software. Still, I don't see it replacing iTunes any time soon--and for that matter, neither does Mufin.

If you want to apply Mufin's similarity engine to iTunes, the company offers a free downloadable widget that plugs into iTunes. Or you could just keep using iTunes' Genius feature, which may not be perfect, but is built right in.