Name That Tune: Audio Fingerprint Firms’ Dueling Business Models

Audible Magic says, "I can name that tune in twenty seconds."

Philips Research one-ups with, "I can name that tune in three seconds."

Those claims are made possible thanks to audio fingerprinting—a highly evolved content-based recognition system whose speed, accuracy, and database of songs is capable of obliterating any mere human contestant on "Name that Tune." For better or worse, however, there aren't plans to stage a Kasparov/Deep Blue-like "Name that Tune" challenge any time soon.

Based on Philips Research's boast, it appears that the Dutch lab is way ahead of its competition. But Audible Magic asserts it has recognition technology that can ID a song in five seconds. The company simply hasn't released it because it doesn't believe there's a single application that demands such speedy recognition. And this is where the audio fingerprinters change their tune; they part ways on the future of audio fingerprinting.

Currently, both Audible Magic and Philips use the technology for broadcast monitoring, which is the process of identifying songs for performing rights organizations to collect royalties for a song's radio play. More recently, audio fingerprinting has become a marketing tool for advertisers to know when their commercials air.

Audible Magic's current revenue stream is from protecting copyright infringement. Its lead product, Replicheck, is a scanning product used by CD manufacturing plants. When a CD manufacturing plant is asked to copy a CD, it first does a Replicheck scan of the CD against 3.6 million tracks to make sure it doesn't contain any copy-written songs. Replicheck is an information tool that helps lower risk. If they were to copy such a disk, they could be liable to the RIAA. Such a case happened to two CD manufacturers. Americ Disc and Cinram both settled with the RIAA for $10 million each.

Philips, on the other hand, has been working hard to get its recognition time down fast. The company believes that with speedier recognition comes the opportunity for new services. For example, when fast content recognition is applied to broadcast monitoring, music pub- lishers and advertisers can be alerted by page to turn on the radio or TV to hear their song or commercial being played.

In addition, Philips plans to launch a new mobile phone application in Europe this summer to identify songs. Simply hold up your cell phone to the radio and in three seconds Philips Research will identify your song. Unlike deploying fast recognition to a known user base of broadcast monitors, consumer desire to identify songs on-the-fly is an unknown quantity. It's unclear what the value of this technology is, which is why Audible Magic still hasn't released its five-second recognition technology. Audible Magic's CEO, Vance Ikezoye says, "The technology is certainly capable of doing that. I think the question we have with that technology is, ‘Can you make any money at it?'"

The audio fingerprint application that's getting all the buzz, thanks to the success of Napster, is peer-to-peer traffic monitoring. Philips thinks the issue is far too complicated because to monitor effectively, one must conduct an infinite rollout; there is no finite user base. Trading happens on a worldwide scale with millions of users exchanging millions of songs. In addition, you have to deal with political issues that vary country by country. Currently, it's a business model that doesn't appeal to Philips. "You restrict people. You don't provide a service. You provide an obstacle. So it's difficult to find a business model," asserts Erik de Ruijter, general manager of the Content Identification Business Unit of Philips Research.

Audible Magic claims it has found a business model. The company monitors activity, but only on a network-by-network level. And it doesn't monitor everything; only songs with the highest trade volume. Audible Magic's P2P Traffic Monitor only scans about 100,000 songs. And even that's too large according to some experts. The overwhelming majority of money being lost lies among a couple thousand tunes. And that database would be easy to create. Just check the top 200 albums at