Mp3 tagging through music fingerprints

I have a collection of few thousand mp3 files with unmeaningful names and tags. Sometimes I do try to rename the tags and filenames using iTunes, Windows Media Player or other editors but its just too huge a task to be completed, more so because the collection is increasing all the time. There is a huge number of tools for fixing, ranging from tools which can edit the tags, to tools which can write the tags using filenames. While learning Python, I wrote a small code to beautify the names, extending which is a Someday task in my GTD list :)

It is too difficult to edit tags using iTunes, even with the functionality of editing multiple files simultaneously. I found this great open source tool Mp3 book helper, which uses the good trick of exporting/importing tags and filenames to CSV files. It is much easier to edit the CSV files using Microsoft Excel, specially because of Macros and functions and then import the corrected file.

However, often the filename and the tags are completely incorrect making it impossible to correct it without listening. I have always wondered if there could be a way to digitize the music patterns and look the tags into a database of music patterns. Today I was searching if there has been some work on it, and found to my suprise that there are many such systems already implemented.

MusicBrainz is such a community which has such a database of music fingerprints and tags. It provides a freely downloadable tool called Picard that can search and correct the tags of your files by sending music fingerprints to MusicBrainz server, although highly manual and not automated. It uses music fingerprinting technology supplied by MusicDNS, which has a database and client side APIs for collecting music fingerprints of various audio files and searching the database. The use of this API is completely free for non commercial purposes. A real act of benevolence I would say, attempting to bring some sort of an order into the huge amount of audio information !

Looks like it can be classified as a Web 2.0 application...the elements are there all right. The database is unique and is constantly improved by users contributing music fingerprints and tags. Programming model is there too, only thing missing is a nifty user interface. It would be nice to write a service which constantly monitors your collection and automatically tags your files by running the MusicDNS api.