For just over a year, I’ve been growing my YouTube channel, which features me playing a collection of American chart toppers, Vietnamese pop classics, and Vietnamese traditional songs on the one-string monochord, “đàn bầu.” The experience has been artistically satisfying, but analytically, the most intriguing aspects have been the insights gained into the nuts-and-bolts of the YouTube platform, particularly its handling of intellectual property and monetization. So for the next six weeks, I’ll be sharing one new insight each week on the YouTube experience from a creator who’s just starting out and putting in his time on the platform.
Here’s a capsule version of the theme you can expect me to cover over the six weeks: YouTube offloads an enormous amount of the labor of sorting through intellectual property law onto its smaller contributors. To be more specific:
- First, its inexact algorithmic tools for identifying copyrighted material mainly benefit global companies with large libraries of content at the expense of individual creators, putting the onus on these creators to prove that they have not infringed rather than on companies to prove that they have.
- Secondly, the messiness of copyright law almost always ends up benefiting corporate interests over the culture of sharing and engagement that YouTube insists it fosters as a company, and people acting in between different regimes of intellectual property (e.g. between “folk” concepts of communal ownership and “legal” concepts of private ownership) are put in a particular bind. This gets increasingly difficult when contending with historical copyright schemes that may not legally apply anymore. For example, music made in war-era South Vietnam was created under a copyright regime that no longer exists, leaving a vacuum concerning their rights.
- Finally, after the recent push by advertisers to have more control over whether their content appears alongside hateful or controversial videos, YouTube has increased its restrictions on what counts as “controversial” which, again, are based on algorithmic analysis that places the burden on creators to convince YouTube otherwise.
Each week, I’ll take a look at these related issues from a different angle arising from my experience releasing videos on YouTube. For this first week we analyze the basic way YouTube’s copyright-checking algorithm Content ID works and observe how it offloads labor to the small-time creators while increasing automation and efficiency for the big industry names.
Setting the Scene
I distribute my videos primarily through my YouTube channel. My presence on other platforms (Twitter, Facebook, my website, etc.) all funnel into the YouTube channel through direct links or video embeds. Accordingly, I decided last year that I would turn on “monetization” — enabling ads on my videos — as a window into the economics of the platform. I did not run the pre-roll ads that are now the larger source of ad revenue on major channels, choosing to stick to the less obtrusive banner ads that appear at the bottom of videos. My goal was not to make money this way (to date, I’ve earned a grand total of $1.17), but rather to participate in the YouTube ad economy to understand the “value” of my videos within the platform.
My channel has largely consisted of cover songs in the past year due to my “Đàn Bầu Remix” series, and my videos can be broken down into the following broad categories:
- Direct covers of pop chart hits with production that mimics the original, with the primary difference being the use of Vietnamese instruments in place of the original vocals, like my covers of John Legend’s “All of Me” or George Michael’s “Careless Whisper.”
- Transformed covers of pop chart hits with production that changes the quality of the song significantly, such as my version of Jason Derulo’s “Swalla” as performed entirely on Vietnamese traditional orchestra instruments, or the really radical re-imagining of “We Don’t Talk Anymore” by Charlie Puth and Selena Gomez as a 1960s Vietnamese bolero.
- Classic Vietnamese pop songs in a style reminiscent of their original versions, like the somber Tết song “Xuân Này Con Không Ve” (This Spring I Can’t Come Home) or the early 1970s rock hit “Mặt Trời Đen” (Black Sun).
- Re-imaginings of Vietnamese pop songs with contemporary-style production, including modern percussion samples and synths, such as “Nửa Vầng Trăng” (Half-Moon) or “Sầu Đông” (Winter’s Melancholy).
- Vietnamese traditional and folk songs, played largely in the traditional style I learned them, like “Ru Con Nam Bộ” (Southern Vietnamese Lullaby).
Of these, only the very last category has no intellectual property problems on YouTube, since “folk music” is considered to be public domain.
YouTube’s Music Policy Guidelines
Due to the range of songs that I’ve chosen to cover, I regularly engage YouTube’s intellectual property management system. First, when I have an idea for a cover song, I immediately head to YouTube’s “Music Policies” page (only accessible by login), and search the database for it. If I’m lucky, the song is in the database and YouTube has already come to some sort of agreement with the copyright holders that allows me to do a song cover, often in return for a share of ad revenue.
Often, the list isn’t up to date, so I must either (1) just upload a video and court the possibility of getting a copyright claim against it or (2) somehow obtain the rights on my own. Since most YouTube creators just starting out are unlikely to have the resources, experience, or desire to do the latter, the most likely options are either taking a gamble or just not doing the song. For example, even though Jason Derulo’s “Swalla” was the featured song for April 2017 MTV Cover of the Month–which I only knew because I entered the contest–it still isn’t in YouTube’s Music Policies database as of May 3, 2017. You could make an educated guess on how a company or artist feels about covers of any particular song based on whether or not other people have uploaded videos, but it still feels like rolling dice.
If the song is actually in the database, then guidelines are usually clear, though sometimes they can be arbitrary. For example, when I began uploading videos to my channel in earnest last year, one of my early covers was of the Usher and Alicia Keys R&B hit “My Boo,” I had not yet discovered the database, so I was just making videos based on whether I saw a lot of other covers of the song on YouTube. I posted the video and shared it on social media. Instead of the excited response I expected, fans and friends quickly stated their confusion: they weren’t permitted to watch it! Yet, I was living in London at the time, and all my European friends insisted they watching just fine. What was going on?
It was through this incident that I discovered the Music Policies page, and here’s what it said for “My Boo”:
Basically, if you wanted to just use the original “My Boo” audio in your video, then your video would be available everywhere but in the U.S. Even where it was allowed, you could not make any ad revenue. Additionally, if you wanted to do a cover song, it would be allowed and you could monetize it, splitting ad revenue with the copyright holders…except in the U.S., where it would not be shown at all.
The Algorithms Taking All Our Jobs
But wait. How was I found out anyway? Is somebody sitting around waiting for me to make a đàn bầu cover of a 2000s R&B song to catch me in the act?
As with much else in our world, this job has been automated.
Some time after a video is uploaded, a system that YouTube calls Content ID determines whether it contains copyrighted material. Though I’m not certain about all the details, I’m guessing that they used the meta-data that one submits (i.e. file name, song title, description, tags, etc.) alongside an algorithmic analysis of the actual audio of the video to match it against audio that is already in their databases. This is probably similar to the kind of audio fingerprinting that mobile apps like Shazam use to recognize a melody. Incidentally, Content ID is used on video as well — I’ve seen bootleggers upload TV dramas to YouTube with the picture stretched, flipped, and otherwise manipulated to thwart recognition algorithms.
If YouTube does find a match, they send you a link detailing the exact moments that their computers think you’re infringing on someone else’s intellectual property. Indeed, the more similar to the original version I make my instrumentals and beats, the easier it is for YouTube’s computers to find a match. On my cover of John Legend’s “Love Me Now”, they found a match for most of the song, but not towards the end, which I assume is when my musical improvisations and slightly differing instrumentation (my taiko drums and string section did not mimic the original sounds exactly) make it harder for the computer to read the audio as a copy of the original. Here’s the readout:
The sheer number of claimants and the fact that they are presented as an alphabet soup of music companies makes the prospect of filing any sort of dispute quite daunting. Where would I even start if wanted to reach some other agreement with then than what they and YouTube had already negotiated? Or if I thought that they were in the wrong?
As a YouTube cover artist, you have little proactive ability to cover your legal bases. The best you can do is read the Music Policies, upload your song, and then wait for the algorithms to spit out whatever YouTube thinks is the appropriate intellectual property agreement based on whatever similarities its algorithms “hear.” In fact, there are a few uploads on my channel which have not been recognized as covers that I would be happy to report to them and just get on with it, but YouTube offers me no way of actually telling them. So, not only has the human burden of figuring out this mess been shifted to me (since on the corporate side, they’re all just relying on the music libraries and Content ID), but I have to do so passively, in response to their algorithms. YouTube’s sound algorithms are essentially negotiating a revenue sharing contract with me on behalf of YouTube and a litany of music corporations.
In some ways, that’s great, because most small-time creators would probably just forego the process altogether and just hope that things don’t go awry and Metallica sues the crap out of them. But as with my rather benign cover of “My Boo,” it can sometimes be quite a hassle that keeps you from sharing your work.
However, if you’re not in it for the cash, there’s one defense that you can possibly use as long as you can make a good case for yourself: the “Fair Use” doctrine, or the basic idea that there are some uses of copyrighted material by third parties that are protected. We’ll delve into that next week, and I’ll share how I eventually got my version of “My Boo” un-banned in the United States.
If you found this insightful, or if you have your own experiences with YouTube’s copyright management system, please share!