Gracenote spins beyond CDs
Slip your favorite CD into your computer. If your CD software works like most, it will almost instantly display the title and artist of the CD, along with the names of each of the tracks.
Most people dont even think about how this happens, and those who do assume that the computer is just reading this information off the CD. But in fact, music CDs dont contain digital liner notes. Most likely, your music software is providing this convenient function using technology from a small company in Berkeley, Calif., called Gracenote, formerly known as CDDB.
Gracenotes specialty is identifying information about CDs and, more recently, MP3 files. The privately held company, incorporated in 1998, began as a hobbyist project in 1995 by some University of California at Berkeley students who were just trying to add a cool feature to their CD software.
The technology of CD recognition isnt rocket science; other companies can easily duplicate it. Even so, Gracenote has firmly established itself as the leader in this niche, with a service that is tough to match.
CDDB, which stands for "CD database," is Gracenotes most valuable asset. Created and updated by its millions of users, the database contains more than 8 million tracks and 800,000 albums. Its technology is used by almost 1,200 digital music players, including RealNetworks RealJukebox, MusicMatchs Jukebox and America Onlines Winamp. Of the major software media players, only Microsoft doesnt license Gracenotes technology, preferring to use a home-grown version instead.
Ty Roberts, Gracenotes chief technology officer, says the company has built up a three-way advantage over would-be competitors: "We have a working system in place, were in all the products, and the database keeps growing and getting richer," he says.
CDDB works like this: Each database entry pairs a CDs title, artist and track titles with a list of numbers, called the table of contents, which is a list of the starting points of each track. The odds of two different CDs having exactly the same TOC is minuscule, so the TOC is essentially a unique identifier for each CD. Nevertheless, it isnt a perfect fingerprint. Popular CDs can have more than a dozen associated TOCs because of idiosyncrasies of the manufacturing process and other factors, and Gracenote must store all of them. Each time Gracenote cant reliably identify a CD, it outputs the closest matches with their associated probabilities. If there are no close matches at all, it asks the user to input the information and adds it to the database. The database gets 20,000 to 30,000 submissions per month, Roberts says.
Gracenotes pricing structure is designed to keep enlarging its installed base of users and, therefore, its database. Any software CD player that makes fewer than 250,000 lookups per month can use Gracenotes service for free, so small developers typically include it in their software. Gracenote charges access fees only when that software becomes commercially successful enough to generate more lookups.
The Post-CD Era
Gracenote is now extending its CD-recognition business with several new digital music initiatives.
Since music files downloaded from the Internet have become popular listening fare, Gracenote has started enabling MP3 recognition in digital music players. But its a task that is more difficult than recognizing CDs, because MP3 files dont have anything resembling a TOC and vary widely in compression and sound quality due to the variety of audio encoders. Gracenotes solution is to use its technology, which it says is part of 98 percent of commercial MP3 encoders, to embed track and artist information directly into MP3 files.
Most CDDB-licensed encoders now include Gracenotes "track unique identifier" in each MP3 they create. They also use a standard format for the text title of the MP3 file, so if tracks dont contain a TUID, Gracenote can often recognize the MP3s from their title. If that fails, the MP3 recognition program resorts to waveform analysis to recognize the music from its melodies. Gracenote says it can correctly identify about 50 percent of the MP3 titles in circulation, and almost always identifies the most popular tracks.
Gracenote also has started licensing its technology to consumer electronics companies such as Kenwood and PortalPlayer, which are creating sophisticated home stereo systems that allow users to encode their CD collections or listen to streaming Internet audio and video.
Another new business for Gracenote, still in its beta stage, is compiling and selling data on the listening habits of the 25 million users who access its database each month. Roberts says this information could be compared with Billboards ratings to better determine the popularity of various albums. With the Gracenote service, "its not just what you have, but how often you play what you have" that gets tracked, he says.
Gracenote Data Services, as the new business is called, will surely draw the watchful eyes of Internet privacy advocates. But Roberts says that unless users opt to be tracked when they download their music players, the company tracks only aggregate listening data.
Whatever shape digital music takes, Gracenote plans to be at its heart. As online music providers -- including Napster move toward fee-based services, Gracenote hopes to provide the technology to track MP3 transfers, so services could divvy up subscription fees to copyright holders without requiring that the music files be encrypted. "Were already at scale with the number of users," Roberts says.