« Over the Edge of the World | Main | Reliable Storage, Redux »

October 4, 2007

Experimenting with MusicBrainz

I have a reasonably sized music collection, somewhere in the neighborhood of 650 CDs. I've ripped most of it to FLAC, and I transcode it to MP3 (using flac2mp3) to load onto my portable jukebox and stream it wirelessly through the house to two Rokus.

The biggest problem I have is getting and maintaining accurate metadata. I've ripped the audio over a fairly long period, and while the ripper I use (SoundJuicer, for the curious) now automatically tags files it creates with MusicBrainz data, it didn't always.

This evening I decided to download Picard and see what I could do about associating MusicBrainz tags with the bulk of my collection.

My first observation is that Picard doesn't like large (9000-ish track) collections - it repeatably locked up trying to import them all at once.

The second observation, once I used the interface a bit, is that even if it had been able to import my whole library at a go, it would have been very difficult to work with - the interface doesn't scale particularly well, and once lists of albums get long, manipulating them becomes cumbersome.

The third observation is that Picard doesn't give me enough control over the tags it writes. All I really want is to add the MusicBrainz artist, album, and track GUIDs to the files - I've spent a reasonable bit of time massaging the rest of the metadata.

For instance - I have a bunch of albums entitled "Greatest Hits" - but when I rip them, I generally transform that title to "So-and-So's Greatest Hits" - makes 'em much easier to differentiate in an album list. The MusicBrainz tagger wants to undo that.

Another example - when a bands name starts with "The", I use it as a suffix - so "The Clash" becomes "Clash, The" - makes them sort under "C" rather than having a big cluster of artists under "T" for no reason other than their love of a definite article. Again, MusicBrainz thinks it's smarter than me, and wants to "fix" my errors.

So I've been tagging artists and albums that don't suffer from these two problems, but even there, Picard falls a bit short. I use [disc #] - to denote individual discs of a multi-disc set, and the data in MusicBrainz seems inconsistent in this regard. It's not particularly surprising for a community policed data set, but annoying none the less.

I've gotten through artists beginning with "D"'s - and I'm not sure how much further I'll go. The quality of the data (and the ability to correct it) makes MusicBrainz superior to CCDB, but it's clearly not (yet?) the panacea I was hoping for.

Posted by dberger at October 4, 2007 9:46 PM

Comments

It's not perfect either, but the tool I've been using to import MusicBrainz data and edit MP3 tags lately is a Java app called Jaikoz. You can get a free 30-day trial.

It's sometimes slow, but gives you complete control over the fields to be imported from MB, do batch editing, etc.

Posted by: Hendel at October 5, 2007 5:02 PM