Robert Kaye has re-launched his MusicBrainz service today. MusicBrainz is set of Free Software tools that are used to fingerprint audio tracks in MP3, WAV, Ogg and other formats, and to create unique identifiers for songs.
What this means is that the MusicBrainz tools can sample a piece of an audiofile, create an "acousitic fingerprint" of the song, and then check with the MusicBrainz server to see if it knows about the song yet. If it does, your music-player will automagically fetch the artist, album, track title and other info (as well as reviews, ratings by people you trust, playlists that include the song…). If it doesn't, you can enter the track info yourself and submit it to the MusicBrainz database so that the next person who comes along will get the info.
This is a lot like GraceNote's proprietary CDDB service — which is how iTunes and other players figure out which CD you have in the drive — but it's way, way better. Organizationally, MusicBrainz is setting itself up as a nonprofit, so there'll be none of CDDB's expensive and restrictive licensing terms for people who want to make players that use the service.
But it's also technologically far superior. CDDB can only recognize CDs. But as music is increasingly distributed online without any CD package, CDDB is getting less and less useful (plus, CDDB is riddled with errors and has a really bad API, so it's hard to build sophisticated services that rely on it). MusicBrainz works off of acoustic fingerprints, which are granular to the level of a single track, recognizably at different sample rates, and work across different file-formats.
It gets better. Because each fingerprint is unique, it means that two people can unambiguously discuss the same track. I can send you a playlist from my computer and your computer can play the songs I'm suggesting, even if you've given them different filename, have them stored in different formats, or have added different metadata about them.
This is also an extremely sweet basis for building collaborative filters atop of. If your computer and my computer can say with confidence that two tracks are the same, we have the basis for collaboratively filtering our collections and finding stuff that we should be listening to — even if we don't know it yet.
There's only one catch: none of this stuff runs under OS X — yet. Which is a goddamned shame, but Robert's broke, and he needs Apple hardware to play with in order to get this stuff ported over to MacOS. This is seriously cool stuff, and all the kids're gonna want it. Let's hope someone out there knows someone at Apple who can intervene on Robert's behalf and get him a loaner so that the Rest of Us aren't left out in the cold.
The answer to this lies in the MusicBrainz community — the community is comprised of individual contributors who work hard to enter and correct the data in the system. The MusicBrainz server software also enforces a peer review system, under which users must review and approve changes made by other users. The peer review system combined with the motivation, expertise and pride of its contributors will ensure that the data in MusicBrainz will be comprehensive and reasonably correct.
Only reasonably correct? No one can guarantee that all the data in a database is correct. Not even the commercial companies that provide metadata services can give this assurance. The MusicBrainz community will respond to problems found in the database and fix mistakes faster than any commercial company with paid contributors can, since the MusicBrainz community is global and is never closed for business. Furthermore, the community is more supportive of MusicBrainz than of other commercial services due to its open nature.
(Thanks, Robert!)
