A Proposal for a Decentralized Youtube Alternative

2019-02-23 (posted in blog)

A decentralized Youtube alternative is an idea that has been slowly percolating in the back of my mind for a while now. A growing theme I see from the handful of content creators that I follow is a general unhappiness with several aspects of how Google runs the service. I am not a Youtuber, but I am a software engineer who enjoys solving problems, and Youtube lately feels to me like a problem that is begging to be solved.

The core idea would be to develop a self-hosted video sharing site that, on its surface, could present very similarly to Youtube. A content creator’s instance would showcase their own videos, could be grouped into playlists, tagged, searched, commented on, liked, disliked, shared, and so on. Instances would communicate with each other using a technology like WebSub, allowing creators to link up with other instances to promote content discovery. A single instance would not be limited to a single creator—for example an instance could host several related creators independently, allowing for networks of related content to share infrastructure and further promote content sharing and discovery.

I envision it as analogous to how Mastodon is a de-centralized alternative to Twitter, only for video sharing.

The Youtube Problems

As I mentioned before, I am neither a content creator nor an advertiser on Youtube. These are some of the problems with Youtube as I perceive them purely through observing the complaints of actual creators. I’m quite certain I could be off-base with some of these, and there may be entire areas of concern of which I am completely ignorant.

Copyright Strikes

Youtube’s copyright strike system errs on the side of the entity making the copyright claim. From what I gather, all that is required is for someone to fill out a form indicating that they are making a legal claim to the content of a video, and the video will either be automatically removed from Youtube or the revenue from that video will go to the claimant instead of its creator. While I do feel that a system like this is important in order to protect actual copyright holders, it is easily abused by people who wish to remove a video from Youtube for non-copyright related reasons. For example if a video is critical of a specific person or company, or expresses an opinion that a person or company disagrees with, that person or company can file a spurious copyright claim and have the video automatically removed.

In addition to strikes generated by manual copyright claims, Youtube has a “Content ID” system that can automatically remove videos that it algorithmically identifies as violating copyright. Presumably it somehow fingerprints background music and video segments to compare against a list of known copyrighted material—I’m admittedly uneducated about its details. This system can be bad for creators in at least two ways. The first is through false-positives that misidentify non-copyrighted material as copyrighted, and the second is removing content that references copyrighted material in a way that adheres to fair-use. Having your video stricken for this is obviously a bad thing, so the more aggressive that Google makes these algorithms, the less likely it is that a content creator might post content that they otherwise would for fear of being incorrectly targeted. This is bad for the creator, because they lose some freedom to post what they want, and bad for viewers because we’re getting a watered-down and sanitized version of what might be very different content if the creator didn’t have to worry about this.

It makes perfect sense for Youtube to make the process of striking videos for copyright violation as easy and aggressive as possible. They have to do this in order to keep the creators accountable for their own content instead of Google assuming that liability themselves. Youtube does provide recourse for creators who have their videos removed unfairly—if a creator disputes a strike, the claimant must provide proof that they are taking legal action or else the video is restored. Despite this, if the strike happened early enough after the video was uploaded, the damage has already been done. Even if the video gets restored, the critical early window in which it would have received most of its views and generated the most revenue for its creator has already likely passed. I’m not entirely certain on the process for disputing Content ID takedowns, but my impression is that there is a manual review process in which it’s also in Google’s best interest to err on the side of leaving the content removed.

In a decentralized model, each creator would be responsible for policing their own content and responding to copyright claims. There would be no central authority taking content offline to protect themselves from liability, and thus no need to dispute bogus claims while you lose views and revenue. An interesting problem that may arise here is how to police content that is promoted and shared on your own instance, but was originally generated by a different creator on an external instance. If that creator were maliciously posting copyrighted material and ignoring copyright claims, it’s entirely possible that the copyrighted material could show up on other instances in the decentralized network through the background WebSub process. A system by which copyright strikes are also propagated between instances could potentially be tampered with by the malicious actor. One solution that comes to mind is to introduce a trusted third party “copyright authority” into the network to handle claims, however this could also potentially be easily circumvented by a malicious actor, and it would essentially be replicating the Youtube problem all over again.

The copyright problem may not be an issue on a decentralized system anyway, so long as each instance maintained responsibility for hosting and displaying only its own content. For example, clicking a “recommended video” from a different instance would link you to that instance, as opposed to embedding the video into the instance from which you clicked it. It might be possible for an instance owner to specify “trusted” external instances for which this behavior could be overridden if a more cohesive and consistent experience is desired.

Demonetization

The problem with demonetization is similar in essence to the copyright strike problem. Google is strongly incentivised to be as aggressive and strict as possible about which videos it deems to be “advertiser friendly.” If Google demonetizes a video from even a large creator, the financial impact to the creator may be significant, but to Google it is a drop in the bucket compared to the impact that a large advertiser pulling its ads from Youtube entirely would have. It’s in Google’s best interest to demonetize videos based on the lowest common denominator. Content that might be considered “friendly” to Youtube advertisers that target a mature audience, but undesirable to other advertisers (think Disney, for example), will likely be demonetized entirely to appease the latter group.

The entire Google-controlled ecosystem is designed to funnel content into being as safe, clean, and uncontroversial as possible in order to minimize the risk of being demonetized. Again, this is bad for creators because they’re producing their content under the constant worry that it might be demonetized. It is bad for viewers because we’re probably getting a lot of sanitized content that’s not true to the creator’s original vision.

Obviously, in a decentralized model demonetization would not be a problem. A creator could be in full control of what ads are inserted into their content, how and when to show them, and could rotate new ad spots in and out at will. Decentralization does impose other problems with monetization in general, which I discuss later in this article.

Toxic Community

Youtube is notorious for having one of the most toxic comments sections on the internet. The bar for entry to get a Youtube account is, by design, extremely low, meaning pretty much anyone who can watch a video can also leave whatever racist, offensive, and vile feedback they want to in its comments. And they often do.

A decentralized system would offer creators the opportunity to cultivate their own unique community. They could set the bar for entry as high or low as they desired, and could have full control through integrated moderation and user management features.

One problem that would need to be solved is how to share accounts between instances. While some instances may desire to keep their users separate and self-contained, there would certainly be cases of networks of instances wanting to share users. For example, instance Y may want to allow a user who registered and logged in through instance X to also leave comments on instance Y videos without having to register or log in to a separate account. There are many possible ways to solve this problem, from a simple shared single-sign-on server to more complicated ones which involve fancy things like blockchain and cryptographic key pairs. I can’t say I’m familiar (or interested) enough right now to go into specific details, but I’m fairly confident it could be accomplished in a way that presents very little friction to users.

Challenge: Advertising and Monetization

Have you heard of NordVPN? Dollar Shave Club? Squarespace? If you watch Youtube (even the paid, supposedly ad-free version) or listen to any podcasts at all, you’ve probably noticed the inundation of ad reads for these and other companies that have started to crop up with an almost alarming frequency. These content-embedded ads are an attempt by these companies to capitalize on the demonetization problem that creators are facing on Youtube. Google can’t remove ads from your video if the ad comes from outside of Google’s ad network and is embedded in the actual content itself. Patreon is another company that undoubtedly profits heavily from Google’s aggressive demonetization policies—creators find it necessary to branch out beyond just Youtube to monetize their content due to the unreliable nature of the platform as a source of revenue.

While embedding ad reads into videos is certainly a way to avoid a complete loss in the face of demonetization, it also (in my personal opinion) somewhat spoils the content. Aside from the obvious annoyance factor of having to listen to a creator blab on for ten minutes about razor blades and butt wipes for the hundredth time every time they put out a new video, it also enshrines the ads into the content in a way that cheapens it. It’s possible for creators to frame the ad read in a way that could easily be edited out later, but unless they specifically took pains to do that, they’ve committed to give ad time to that company in perpetuity. Even if NordVPN shuts down tomorrow, or was bought out by a company who decided to no longer respect whatever affiliate deals were previously in place, there would still be thousands, maybe millions of pieces of old content being listened to every day singing the praises of NordVPN and earning the creator absolutely no money. Now, I can appreciate the appeal of old commercials in time. Watching old TV ads from the 80s and 90s is actually something I occasionally do for entertainment purposes. And I run a streaming Old Time Radio service where having certain old ads included in the episodes can actually be more desirable than clipping them out. But somehow I don’t think a Squarespace ad read will have the same endearing qualities in 60 years that a Lucky Strikes musical jingle from the 1940s has today—and even then, you can only hear “light up a Lucky” so many times before even that begins to grate.

A decentralized replacement to Youtube would give creators far more control over how they monetize their content. They could record their ad reads separately and inject, rotate, and remove them from their content at will without having to re-edit or re-upload anything. Google’s ad network would essentially be replaced by a network that the creator controls. The creator would no longer necessarily have to cater their content to skittish advertising partners like Disney because they could simply not pursue ad relations with Disney in the first place. Functionality similar to Patreon could be baked right into the software—allowing instance owners to plug into payment processing back-ends to offer their users the ability to donate, subscribe, purchase merchandise, or otherwise contribute directly to the creator in exchange for special (optionally tiered) access to otherwise hidden content.

The true problem would arise with finding sponsors or advertisers in the first place. The appeal of Youtube over self-hosting is most apparent in that it’s owned by Google—a multi-billion dollar advertising company that advertisers are more than willing to do business with. They are a one-stop-shop for distributing video ads, web display ads, text ads, basically every kind of ad imaginable. They gather obscene amounts of data and statistics on audiences, package it up and sell it back to the advertisers who use that information to buy and place even more ads, doing million of dollars of business all through the same interface. To think that such an advertiser would also branch off and do direct business with Joe Schmoe’s Video Sharing Site that gets a few thousand views a day is absurd. It’s a problem that would probably solve itself the same way that web advertising was solved in the early days, leading up to what we have today. Video ad networks would crop up and serve commercials from several partnered advertisers, and instances on the decentralized network could tap into them for a share of the revenue. These services would compete with one another, serve different niches, buy each other out, and eventually conglomerate into a new Google-like entity that essentially has a stranglehold over the market. It sucks, but it might be a fun journey on the way there.

Challenge: Convenience, Discoverability, and Ubiquity

This is the actual reason that a decentralized video sharing network could most likely never be successful, relegating this entire idea into a theoretical problem-solving exercise as opposed to an actual viable project worth spending resources on.

Youtube is literally everywhere. Pretty much every entertainment-related technology under the sun supports Youtube out of the box. You can play it on your smart TVs, Rokus, various streaming sticks, video game consoles, phones—heck, I bet my watch could even play Youtube videos if I sideloaded the right Android apps onto it. Your phone will also notify you every time your favorite creators post a new video or start a new livestream, and this happens (again, by design) almost magically without you having to set anything up or even ask for it. No decentralized system could even come close to that level of convenience and automation. The fact that Youtube is so ubiquitous means that the size of your potential audience on any self-hosted solution, no matter how popular or how many other instances you partnered with, would almost certainly be a miniscule fraction of your potential audience on Youtube. For a viewer to enter the Youtube ecosystem requires essentially no effort at all. In fact, at least in the case of Android devices, it would take a concerted effort on the user’s part to not find themselves forcibly inserted into the Youtube ecosystem.

In my old life working in the advertising industry, the conventional wisdom was that anything in the “funnel” that requires the user to take an action immediately loses you at least 90% of your users. Do you have an opt-in pop up before you install your malware on a user’s device? Turn that into an opt-out instead and you now have 10 times more users getting past that step than you did before. In order to entice adoption of any new platform, the process of acquiring and using it needs to be buttery smooth and require as little user interaction as possible. A decentralized video network would essentially lose that battle out of the gate. It would be competing against Google, who owns and pre-loads Youtube onto the very operating system that any competitor would need to gain traction on in order to succeed. Being decentralized, there would be no easy or obvious point of entry for smart TV or streaming device manufacturers to pre-bake into their offerings. While it would certainly be technically possible to support such a network on those devices (assuming it defined and adhered to some set of standards), it would likely require the user to already be aware of the feature and know a specific creator or instance to plug in.

In the end, while I wish we lived in a world where a scrappy young engineer or team of engineers could whip up a decentralized, creator-friendly, self-hosted alternative to Youtube that could actually compete, the reality is that we probably don’t. We, as a society, have allowed companies like Google, Apple, and Amazon to gain such a stranglehold over the technology that we use day to day that extricating ourselves from their web is already extremely difficult, and will probably only get worse in time. Projects like Mastodon and Purism do give me a glimmer of hope, and I desperately wish to be proved wrong and will always support those kinds of efforts, but it’s hard to stay optimistic when paying attention to the direction that technology is heading these days.

All that pessimism aside, if anyone out there is actually working on or considering creating a decentralized Youtube alternative, please ignore everything I just said, and best of luck to you!

Short Permalink for Attribution: rdsm.ca/3vxqj

Comments