Obviously Lemmy, Mastodon, Pixelfed, etc. are federated decentralized equivalent to their centralized counterparts, but what is the counterpart in the fediverse to TikTok? It is a dominant app for millions of people, and as far as I can tell the closest thing is Peertube, but isn’t that more of a YouTube equivalent? Does it not exist because the bandwidth and storage costs are just too great? Or because the algorithmic nature of content selection is inherently anti-fediverse in some way? Clearly many people choose to interact with each other this way, but it seems like a gap in the fediverse and I was wondering why.
Non-retardation?
TikTok works because it has a powerful algorithm and because people like being blasted with non stop shorts. That’s really not healthy at all.
I have had an idea for doing this for a few years now, the only issue video storage is costly
I’m not a Tiktok user, but what does Tiktok do that something like Peertube can’t, and that we would actually want as part of the Fediverse?
There really isn’t one.
As far as I know, no one tried to clone TikTok for the fediverse. But I think the inherent problem is the algorithm. TikTok isn’t like Youtube or any other social network, where you follow people. You have an algorithm that tracks everything you do and watch and then suggests you video based on their topics, less on the people in them. I guess it’d be hard to implement, as many in the fediverse are not keen about tracking.
But what all is needed for tracking? AI-based metadata extraction from the video and metrics of how long the user watched the video before swiping or rewinding or stuff? That can all be done at the instance level. I’d imagine the bigger issues is the engineering involved in the app content creation tools, and the costs of data storage for all of those videos and bandwidth for distributing them, something TikTok currently just foots the bills for. Arguably an open source equivalent wouldn’t have the privacy stigma around it, right?
Tiktok takes into account much more than that. Have you looked at the app permissions? Location, sex, age, device info, data from 3rd parties, camera, microphone, clipboard, contacts… And that’s just off the top of my head. Hiding that fact makes it very convenient because I sure as hell would not voluntarily give all that data (and more) to anyone.
Honestly, I think the cost of data storage and bandwidth would easily be the biggest hurdle. Video takes up loads more space than text or even images, so I’m not sure if it would be feasible for any volunteer entity to support unlimited free video uploading.
That’s where smaller, niche hosting sites could pop up and mesh together through federation. Sure, big corporate hosting sites will exist, but many content creators will probably turn to smaller instances or even self host, so that they can have finer control of their videos, communities, and profits, but still be compatible with the fediverse and thus compatible with federated video players. It really puts power in the content creators hands. I am interested to see where it goes.
Conceivably you could open source the algorithm, or even better, have a variety of algorithms to choose from with custom parameters.
In a similar vein, I’m not sure if anyone remembers Slacker Radio, but it was a competitor to Pandora/Spotify/etc. It had its drawbacks (hence why it isn’t around anymore), but I absolutely loved the amount of control you had when building custom stations. You’d first seed a custom station with a bunch of musicians you like, and then there were a number of parameters which allowed you to fine-tune the algorithm to a remarkable extent, well beyond what today’s music apps offer.
I’d love to get to a place where we have options other than just saying “welp the algorithm” and just giving up, I think that the ability to customize one’s algos would be a killer feature that the fediverse can offer which the major platforms generally won’t.
You could probably run such an algorithm locally by logging data locally then determining a type of video which could then be requested from the server. On top of that to get even more data privacy you could store the data collected for this as different numbers in a 3d array or other data structure and then use some form of math to determine what would be best based on that so any personally identifiable information will be really obfuscated and nearly impossible to use for anything other than the algorithm it’s designed to be used by. This would then allow users to not only fine tune their own personal algorithm parameters, but could open the door for allowing people to write their own algorithms or systems to do this.
might be a good idea for someone to make such a thing as TikTok has been getting closer and closer to getting banned in more and more places. There may very well be a twitter-like opportunity for someone with such an app to swoop in and gain some market share
The inherent problem is money. Sites that store and serve text can be very cheaply run. Sites that store and serve video are expensive. The storage and throughput demands are much higher. In order to get videos to load quickly, you need a CDN. Nobody of average means can run a TikTok clone as a hobby.
