Nicholas Pell
May 23, 2012

Google DRM software threatens social media

In their Youtube video, @KarminMusic covers Katie Perry's "Fireworks." With Google's new melody-matching technology, covers such as this could be identified and removed. The RIAA has always been a little twitchy about music copyrights, and that’s putting it mildly. The latest front in the war on piracy (or fair use, depending on where you stand) is YouTube. Filled with album cuts, live versions and amateur covers, it’s become a stronghold of sharing media without file sharing. Needless to say, this isn’t the record industry’s favorite practice. But while technology exists to catch the album version of a song or an official live release, there’s nothing that can prevent you from uploading an iPhone video from the last concert you went to or even recording your own version of a song without permission from the songwriter.


Google, however, is looking to change all that with its new Melody Identification DRM. The Mountain View-based company recently applied for a patent that will (allegedly) be able to pull a melody out of a song, allowing YouTube to recognize when your 14-year-old daughter has uploaded a video of herself singing the latest Katy Perry jam. The goal is to help the record industry crack down on copyright infringement that is currently difficult to detect.


Here’s how it works: The software takes a sample of the audio track on a video. However, this sample is different from the samples used in DRM currently. This is because the program isn’t trying to match the exact recording, but rather the actual melody of the song. The software does this by comparing the pitch at different points in time throughout the track to a master copy kept by the company. Too similar a pitch at too many times over a long enough time line and digital copyright protection kicks in. This could take many forms, from removing the video in question to slapping a bunch of ads on top of it.


The software in question can use the standard, Western 12-tone scale or it can do something a little more nuanced, like a 36-tone scale to really make sure it’s doing the job right. The difference between melody recognition and sound recognition can be likened to the development of “talkies” at the end of the silent film era -- a potentially game-changing quantum leap over and above previously existing technologies. It can also determine who owns the copyright, or if the person who has posted the offending video owns a portion of the copyright.


If you’re skeptical that this will actually work, you aren’t alone. The RIAA must have been working on such a program for years, though they’ve yet to produce one. Further, while the nuanced ability of the program to pick out a song are no doubt impressive, one can’t help but wonder what types of Kafkaesque SNAFUs are going to result from delegating so much work to a computer. Especially while the kinks are getting worked out, it doesn’t seem unreasonable to think that tons of songs will be illegitimately removed from YouTube and other media hosting services because they sound too much like a live version of another song.


One example: The 1-4-5 chord progression is the cornerstone of early rock, tradition R&B and punk rock. It’s the chord progression heard in everything from “La Bamba” to “Louie, Louie” to “Blitzkrieg Bop.” These songs -- while great -- aren’t precisely known for their musical nuance.
There’s also the small issue of parody. The Supreme Court has ruled that musical parody is speech, not commerce. What this means is that there’s nothing stopping you from taking the hot new Lady Gaga track, recording your own version about how much you don’t like Lady Gaga and then posting it on YouTube. This is, in a world where the scope of such has become increasingly narrowed, “fair use.”


Further complicating matters are bands like the Grateful Dead, who don’t care if people post shows recorded as bootlegs from the audience, but prefer to maintain control over concerts recorded straight from a sound board. Some copyright holders, such as Greg Ginn of Black Flag and the Zappa Estate aren’t opposed to people posting their music on YouTube per se, they just want it to be of high audio quality and without advertisements or personal videos. Determining the correct audio quality seems easy enough. However, how will this DRM software do anything to tell if the video is an image of the album cover or grainy shots of kids skateboarding?


While the recording industry is desperately looking for a way to crack down on sharing music on the Internet, they’ve yet to come up with a good way to do it -- other than getting people to sign up for Spotify. Google’s recent patent application doesn’t seem to change that.