Near-Magical Music Identification

By Dave on September 15, 2008

Quite by accident, I discovered something new about iTunes today; it can identify music on a CD that you burn from vinyl, seemingly by way of magic.

Over the weekend I was talking with a friend about some of the music we listened to years ago, and I offered to lend him some that I thought he’d enjoy. I was digging through some of my old LP’s and cassette tapes yesterday, and on a whim decided to digitize a few of the albums and put them on CD — he doesn’t have a turntable, and it’d been a while since I’d listened to them & thought it’d be nice to listen in the car.

I used one of those CD’s today — not exactly CD-quality, recorded as it was from an album that was cut in 1981 — but it worked great (and it’s almost scary to realize that I still knew the lyrics word for word!) I brought the CD in the house this evening, and my son was wondering what it sounded like, so I popped it into my wife’s kitchen computer (an old Mac laptop) so we could listen. Much to our amazement, once iTunes fired up, the album name and song titles were right there!

Now keep in mind that this CD started out from a dusty 27-year-old LP, played on a not-very-well-kept turntable. The turntable is connected to a receiver, which has a patch cable with an 1/8” stereo plug connected to the mic input on my computer; pure analog. The analog audio was digitized using a shareware audio editing app (Amadeus II) and saved out to a plain vanilla aiff file, then burned to a CD-R with Toast. I didn’t go to the trouble of naming any of the tracks, nor did I do anything to help the computer identify the music, much less take the time to clean up the LP’s hiss & pop from the audio. But in the time it took to launch iTunes on a 233MHz G3 PowerBook, it was able to sample the music on the CD, connect with Gracenote CDDB and match those samples to the entry in their database. Wow.

The magic happens via the Gracenote Recognition service:

As a recognition service, we… enable… applications to access our database and recognition services. On most commercial music compact discs, there is no disc title, artist, track title, credits, or other information that a player running on your computer can extract and display for you.
… However, a licensed application can lookup the CD in the database and return the title and track information to the application for display by the time you hear the first notes of the album.

I understand the basic mechanics of what’s going on, but it just floors me that it can happen so fast. I’ve seen iTunes identify CD’s before, but they were all commercial disks, and just figured that there was something digitally encoded on the disk that helped iTunes and Gracenote know what was on it. I’ve also used my Verizon Razr V3’s V-Cast Song ID app a number of times, but that has to “listen” for 10 seconds, then has to think about it a little before it can identify a song. But iTunes did an entire CD in a matter of three or four seconds.

While it’s not exactly magic, the way it works is still pretty magical, don’t you think?

###

Comments

  1. Darren says:

    And what a classic album (and artist) to bring this magic to light.

    As a question, does anyone know if Gracenote uses the number of tracks / track lengths to work some of it's magic, or something like that?

  2. Rick says:

    Gee, Dave - does this mean I'll be able to enjoy the magic this weekend? :)

    Looking forward to hearing some good stuff - many thanks!

  3. Dave says:

    You got it, Big Rick! You could even stop by earlier than that if you're in the neighborhood.

    Darren: Not sure if the track length would be accurate enough. The length of a given song is somewhat variable, especially when dealing with an LP; a little variation in the turntable speed could vary the length of a song by quite a bit, then you throw in manual editing to trim the dead air from the beginning and end... But then again, if it matched a sound pattern on the first track and then checked the lengths of the others to see if they fit within a certain range, that might work. But that's obviously just a guess.

    Glad to hear I'm not the only one who appreciates some Larry Norman!

  4. benxamin says:

    I'll check-out Amadaues II. I use the Final Vinyl 1.0 (that comes with the iMic) to import into Quicktime for editing. Then export to .aiff. When I drag the .aiff files iTunes, they behave normally. However, they will not play on the iPod. Kind of defeats my purpose for digitizing a record collection.

  5. Dave says:

    ben: Amadeus is nice for editing, but Audacity works pretty well too and is an open-source freebie.

    I think to get the aiff files to play on an iPod you need to convert them to AAC or mp3 first (set it up in Preferences - Advanced - Importing, then right-click on an aiff file in iTunes and pull down to Convert to...). aiff files should play on an iPod, but I've heard of others having trouble with them too. Besides, putting uncompressed audio on an iPod chews up a lot of space in a hurry, and most people won't hear the difference between the two. Definitely keep the high-quality aiff and/or burn it to an audio CD, but use compressed files on the iPod.

  6. benxamin says:

    Thank you, Dave! I just downloaded them both and will try them out.

  7. Peter says:

    "Hold the phone up, record a sample, send it through your network provider, and GraceNote will match it up with its acoustic fingerprinting database." - (click on my name for the URL) How cool is that? CDDB only generates a check number (ID) based on the track layout on your CD and relies on the assumption that no two CDs with different sounding files could ever have the same exact layout. GraceNote is much cooler!

  8. Peter says:

    anyway.. my point being it works for any source, even a cruddy 3KHz sample from a cell phone microphone... unbelievable.

  9. M E-L says:

    MusicBrainz does this too -- basically creates a digital fingerprint of your mp3 and compares it to its user-created database.

  10. DavidR says:

    Gracenote CDDB does indeed use the track lengths to identify the album. It was just lucky that the signal levels detected as start/stop points for each track gave timings that were close enough to those calculated by the program that you used. Or they could even have been uploaded by another user who had ripped the LP exactly as you had done...

  11. Dave says:

    Lucky? It's worked on half a dozen other LP's I've digitized since writing this. Seems more like magic to me. ;)

Comments are closed. If you have something you really want to say, email editors@gadgetopia.com and we'll get it added for you.