Both Peter Kafka and Lyor Cohen seem to be mystified by concerns about how YouTube shakes down artists and songwriters in Kafka’s interview with Lyor in Recode. It should come as no shock that Google chose Kafka (a long-time Google fan boy) as the safe interviewer for Lyor, but it really is stunning how oblivious Lyor is to how to scam Content ID which has leaked like the proverbial sieve since its inception. And unlike Lyor, I’m just not willing to believe that the Content ID team at Google doesn’t know exactly what they are doing. Or not doing.
When I told [Google’s Content ID developers] about the albums, they said, “Yeah, they jumped over our Content ID by speeding up the tempo of the music, slightly. We’ve already got a solution for it.” I had them walk me through the process. I felt so proud that I could really talk to people in the industry that had this feeling about Content ID, and finding bad actors, and confidently say, “We’ve got a team that is dedicated to fixing this.”
The more we frustrate bad actors, the more we can stop cottage industries. They’re just going to give up, at a certain point. And I think the industry will feel really good about that.
Couple facts: Content ID works by comparing the soundtrack of a YouTube video to an audio fingerprint reference of a known recording. Let’s be clear–the Navy has used psychoacoustic fingerprinting since World War II–essentially the same technology as Content ID. (Find a contemporary research paper on fingerprinting here.) Remember Jonesey in the Hunt For Red October. They called it “sonar” in the movie, but it was actually a method for capturing a sound in the wild and comparing it to a known sound.
Jonesey was able to identify Red October by manipulating the speed to create a new fingerprint with which to track the Soviet submarine. Red October was able to fool the U.S. Navy acoustic fingerprints by an new engine type that made a sound the service had not recorded before. Jones manipulated the speed to determine if the sound the database identified as a biologic was actually man made. Once identified by the reference copy, the database could be programmed to recognize the sound captured in the wild repeatedly as long as it did not vary from the reference version.
The acoustic fingerprint is a mathematical rendering of the wave form of a sound. Machines then can look for what is essentially a number and when they find a match, take some preprogrammed action depending on the network the machine is connected to.
If you were on the side of the good guys, what you would not do in this situation is offer an easy way to defeat the 1940s technology by changing the speed of the sound captured in the wild. Unless, of course, the plan was to use speed as a countermeasure to defeat detection.
And it’s not just music that is getting manipulated, it’s movies, too. Like this illegal and monetized copy of the movie Jarhead that was easy to find using the “full movie” search term for either Jake Gyllenhaal or Jarhead:
The YouTube partner is slowing down the monetized illegal copy of the movie by 25% TO DEFEAT CONTENT ID. The YouTube partner then puts a little bubble in the beginning of the movie telling YouTubers (who know to look for such a cue) to speed it up to 125% of the upload speed to normalize the movie. (A method that has the added benefit of defeating anything looking for running time, but that’s another story.)
What is the necessary factor in making this kind of piracy work? Easy–the YouTube viewer speed control. A speed control that is common to editing software but is unique to YouTube among commercial user interfaces. Why is it there? To cater to fans of Alvin and the Chipmunks?
No. It’s there so YouTube can monetize illegal copies of music and movies.
Even if the music is edited into a video that is intended to be viewed at normal speed, changing the pitch (aka “pitch bending”) may defeat Content ID if the music used is itself speeded up. There are crude and sophisticated ways to do this, but the intent is the same.
And there are dozens and dozens of “how to” videos on YouTube that tell users exactly how it’s done. Like this one:
This particular “how to” also has advice about how to defeat Google’s “copyright strike” system by always appealing a block because no one is going to sue if you send a counter notification. Good for users, good for Google, money flows to everyone except the artist or songwriter. I’d suggest this is a prime example of the “DMCA license” that is neither a license nor DMCA compliant, but uses counter notification as a sword.
If Google were serious about piracy, they’d dump the speed control on YouTube. They’d also police the “how to” defeat Content ID videos on YouTube. If they want to keep their speed control and have these videos on YouTube, fine–just don’t get all sanctimonious about how serious Google is about defeating piracy while at the same time using these tools to gain leverage when making a commercial deal. It’s really one or the other. And if Google were serious, they’d also update the woefully inadequate repeat infringer policy.
Maybe not as important as stopping Google from sharing advertising revenue with ISIS or stopping Backpage from profiting from selling children, but since Lyor brought it up….realize it’s all bullshit.
And it’s not just me–HITS says so, too. So it must be true.