153 points | by aestetix3 days ago
Focusing on video game material instead of being neutral and coming up with a reasonable business model that makes sense for all your customers (then communicating it up front) is the problem. There's always going to be a subset of customers that pushes the envelope. This conflicts with short term growth strategies but perhaps there's a little room for ethics to sneak in.
https://www.cnet.com/tech/services-and-software/vimeo-bans-v...
Are you implying that the videos getting hit by this are not generally video game content?
I don't think that's right. In particular there's a lot of speedrun recordings that are going to disappear when this day hits.
(The HN title had to cut some of the explanation for brevity, probably 99% of what is going to get deleted is long highlights, not uploads.)
https://www.sec.gov/Archives/edgar/data/1652044/000165204425...
You're jumping to the assumption that surely YouTube's costs have to be lower than $36B, and that is not at all assured. They handle an absolutely gargantuan amount of network data transfer, not to mention processing compute. I'm ignoring the storage but even that at their scale is probably at least 1B.
"From Q4 2023 to Q3 2024, YouTube's combined revenue from advertising and subscriptions exceeded $50 billion."
Yes, network and compute is expensive, but when you are the size of Google the economics look a little different.
It seems like you just have to be sufficiently large before you can successfully monetize a video platform.
Edit: others have explained elsewhere VODs are auto deleted after 60 days, and then must be converted to highlights, which will be affected. I think anyone who relies on Twitch VODs as a viewer or producer is a glutton for punishment anyway. The viewing experience is dreadful if I remember correctly, enough so I just wait for a YouTube upload anyway.
In my anecdotal experience, I have probably watched several thousands of hours of live content over the last decade-plus, and maybe a half dozen VODs.
What do you think is dreadful about it?
To me the problems center mostly around their site and more so the mobile app seem to be designed to discourage watching VODs as much as possible, you really have to dig down to resume watching something you started watching yesterday.
The place where twitch really lacks isn't on vods, it's the fact that the live player can't pause or rewind at all.
...Those that are of interest anyway, which turned out to be less than 1% of what I thought. :D
Twitch's player is indeed terrible and _loves_ to forget where you are. It's hard to believe this is still a problem in 2025...
We would be at a penny per terabyte of space. If AV1 in HD can store 400 hours of video per TB, the roughly 24TB to store a 24/7 stream over the course of a year would cost only 25 cents. Providers could keep all video content indefinitely.
Perhaps there's some benefit to this exponential growth coming to an end. Imagine a surveillance state that had near limitless storage and could keep 24/7 recordings indefinitely of cameras on every street, house, vehicle, etc.
And keep in mind that security footage has an awful lot of both static and repetitive objects. It's an ideal target for more complicated approaches to compression.
[0] https://github.com/ppingzhang/Awesome-Deep-Learning-Based-Vi...
Thank you, your comment went straight to favorites.
The word you're looking for is "frugality." This is a tail use case that isn't cheap to run.
> https://www.aboutamazon.com/about-us/leadership-principles
Leadership principles? Are you saying some exec / director ("leader") is looking for a P&L bump for their promo doc?
This affects far more people at a much higher scale than Twitch will admit, and the deadline given isn't enough for these data transfers to complete.
Vod viewing on twitch is also a pain, ads every 10 minutes, buggy playback, and vods don't play in order.
What's going to happen is anyone currently storing their playthroughs on twitch is now going to export to youtube. So I guess they want youtube to get the ad rev.
So you'd hit the limit after 600 ten-minute videos, or 100 hour-long videos.
The limit also seems to apply to "Highlights" and "Uploads" but not to "Past Broadcasts", "VODs" or "Clips" for added confusion.
One might imagine this is just the logical followup of them adding that horizon initially, basically saying "the 1 in 200 of you who circumvented our policy, no, for real, stop that."
In this case, they seem to be saying long-form archives aren't helping their business and are very expensive.
Of course, since that also de facto means people start pointing to their YouTube pages as their content archives, that means they think they have such a better platform for live content that they can survive people doing the calculus of "well, if I have to host my old content on YT anyway, why am I using Twitch if I'm just going to upload to YT after..."
Whether that's true or not, we'll see. (One might argue this is a given comparing the number of people I know who stream on Twitch versus YT, but Twitch is also the place that thought people wanted them to integrate a game store in their desktop app, and appears to have the attention span of a squirrel in long-term platform initiatives, so...we'll see.)
(I work for Google, I've never worked on anything related to YouTube, opinions my own.)
In practice though I doubt this makes a huge difference either way, the vast majority of the people that can have noticeable amount of views on such already have their YouTube channels or other venues they are also making money from.
> This won’t apply to Past Broadcasts (VODs) or clips.
If you see a video-on-demand that is older than that, then that is an “upload” and not a “VOD” and thus is in-scope.
So yes, this will absolutely affect the speedrunning community, and anyone else who has been using this method to archive old streams.
I was under the impression that the principal objective of speed running was to get things done quickly. You should be able to fit a lot of valuable information within the quota if you are any good at it.
It's not merely doing something quickly; it's more akin to a sport.
The objective of speedrunning is to perform something you would do in a game in a record time, or it's now been somewhat expanded to sometimes include or mean some extraordinary feat in a game that may not be directly related to speed.
A speedrun of a game might mean to complete a game that would normally take months in (for example) "only 10 hours", in which case the speedrunner needs to be live for those ten hours. A recording is not an acceptable substitute due to issues of cheating[1].
Even if a speedrun is only two hours, a speedrunner may need to play the same game four, five, or twelve times in order to achieve their objective. They could be playing for an hour and fifty minutes only to have the entire run ruined by a mistake, or even just a random game event.
[1] It's still possible to cheat live, but it's more complicated, more challenging, and there's a greater likelihood of being caught.
I am still not following why Twitch needs to maintain live copies of all the failed runs. Once you hit the objective, make that video the highlight or whatever to be persisted indefinitely.
Why would anyone care about watching several hours of something when they know ahead of time it's not going to be representative of a successful outcome? Iteration #17 out of hundreds can't possibly be valuable enough to justify the storage cost in even the most charitable of cases. It seems to me that most of speed running could be done completely offline without involving the internet and video capture technology (i.e., practicing a musical instrument).
But part of the reason this has become such a popular thing is the community aspect of it - people get drawn in and inspired to participate because they get engaged in the community of either particular runners or the wider community of people who follow all the runners of some games.
At least for me, while I've never had the desire to participate, when I was sick for a year or so, and therefore at home with little ability to participate in a lot of other things, I went down the rabbit hole of watching different runs of different games, and one of the more useful tools and timesinks was being able to watch the past broadcasts of different runners and seeing if they were enjoyable to watch, at the particular game whose speedruns were interesting me at the moment.
And since not everyone just runs one or two things, sometimes their last runs of those games were months in the past.
So at least in my n=1 experience, those broadcast archives specifically were quite useful for me as a viewer and person attempting to discover more streamers to watch.
Streaming games has a large social component, whether it's speedrunning, or just casual play. It's often as much about the personality of the player as it is about the game. People watch as a communal activity.
https://www.youtube.com/watch?v=FezT4GTZj0g
:)
FAQ why its so long
- Why is it so long? In this game, items evolve over time. There is an item, the Shampoo, that takes literally 2 weeks (= 336 hours) to evolve into Splendid Hair.
Speedrunners are often playing the game or parts of the game hundreds of times. And they're usually performing techniques that take lots of precision and therefore lots of practice.
So they stream it all, documenting their attempts and trying new strategies in front of a live audience. They produce so much comment that there are YouTube channels that make documentaries about different speedrunners.
You may be thinking of TAS (Tool Assisted Speedruns) which is a separate thing.
[0] https://www.microsoft.com/en-us/research/blog/introducing-mu...
Not to mention the huge amount of voice samples and webcam footage you could use for more typical voice cloning, text to speech, human avatar creation, etc
Past Broadcasts (VODs) and Clips are unaffected.
Making a VOD a highlight was a way around that -- Twitch would never delete those.
I'm not disagreeing that it was common knowledge that this was a way for non-partners to circumvent the regular retention policies, which is why this 100-hour limit seems like a pretty generous compromise.
Clicking through random channels just now, I didn't see a single account with any Uploads, and most of the channels who had any content in Highlights seemed to use it pretty sparingly (<100 hours). It doesn't seem like a common practice, and Twitch doesn't seem like it's trying to eradicate history, just reign in some behavior that the platform didn't intend to support.
If you're aware of certain communities who've made a practice of highlighting their entire streams (beyond 100 hours) without being partnered, maybe you could promote them here so people could help archive them?
It turns out that most people who cared enough about retaining VODs just upload them to Youtube. Youtube is simply a better viewing experience for non-live videos, and it can generate some revenue (though usually very small, unless you have a huge amount of views). One problem with Youtube is that it's more strict about (what it thinks is) copyright content -- for example, some otherwise "free"[1] videogame music is regularly claimed by on Youtube by someone who sampled the original song, so it registers as someone else's content to Youtube's content ID.
[1] By "free" I mean that original videogame music is not usually actively protected (even though it's under copyright), because publishers love when people promote their games.
This VOD is over 2 years old [https://www.twitch.tv/videos/1891768073](https://www.twitch.tv/videos/1891768073)
Automatically save broadcasts for up to 7 days (14 days for Affiliates, 60 days for Partners, Turbo and Prime users)
Hosting huge amounts of video data throughout multiple data centers is just plain expensive. Some ads isn't gonna fix that, especially since twitch already has a frustrating amount of ads everywhere.
https://bsky.app/profile/authorblu.es/post/3likxmdytys2l
Assuming ~6000kbit/sec that's about 17TB of archived video for that guy alone.
It's possible the actual additional storage requirements for that specific user are minuscule, since we don't know what data they are/aren't archiving themselves, if they're doing any deduplicating, etc.
Billing the owner a few bucks each month the each thousand hours of extra storage would make much more sense than removing everything.
That's the cost for just buying disks, but storing data in the cloud costs more than that and it's an ongoing cost.
S3 charges 1.25c/GB/month for this sort of data. So that's $200/month for just this guy. There may be 100s or thousands of these people. Easily adds up.
> S3 charges 1.25c/GB/month for this sort of data.
It doesn't cost them anywhere close that. Their competitors charge twice as less or more an still make money.
Twitch belongs to Amazon, they are the cloud.
Setting up your own infra to handle this is of course going to cost you a lot more than that, but when you have the infra set up then the marginal price is hardware (+ a monthly electricity bill, which is not as high as for other kind of workload).
And even if they had to charge $200 a month, they should probably offer the option instead of just removing the content: we're talking about professionals who make money out of the platform (and earn Twitch their income), they can make the choice whether or not they can afford it.
There's no way these professionals have 6000 hours of interesting content and there's no way they would pay $200/month to store it. They're just saving everything they ever record because it's free.
Implementing that feature would cost more money than it would ever make.
That's all just off the top of my head, and all of that is going to be fighting against all the other projects that people want to get done, projects that are likely way more profitable and way closer to the primary goal of the company -- being an intentional streaming service, not an accidental video hosting service.
If they annoy the most active streamers to the point they leave to another site, why should a viewer stay at Twitch versus just using another site?
I’m assuming some of these accounts bring in far more than the $500-1000 it costs to host old video.
Going from an unknown limit down to 100 hours with little notice shows how shortsighted Twitch was here.
You’re not entirely wrong but you’re exaggerating the difficulty.
6Mbps is Twitches recommended ingest bitrate, and their highest quality just serves the ingested stream back to viewers without transcoding. In reality the storage would actually be a little higher still because they have to store all the transcoded lower resolution versions as well.
Interesting, I wouldn't have guessed that it would make sense for them with regard to bandwidth cost. TIL, thanks.
Twitch will offer a premium sub for heavy users most likely.
Is this content searchable in any meaningful way for the client?
Btw, I think there is an easy option to export your Twitch content to YouTube so that's another way of saving all the content.
When you think about social audio, who is number one? Spotify is music subscription service and it's not really YouTube for audio meant for content creators and SoundCloud is stuck in time and it never really took off.
I would like to see SoundCloud reimagined with new features and ideas.
X Spaces is also big-ish (?), but mostly used by crypto scammers.
https://professionalartistmag.com/how-film-saved-now-infamou...
Run it yourself, and you can save whatever videos your community cares about for as long as you fucking feel like because YOU OWN IT. None of YouTube's asinine copyright strike bullshit to worry about -- if a company has a problem with your use of something they need to send you a real DMCA notice. None of Twitch's random policy change bullshit to worry about. No advertising. If your community actually gives a shit about the content then they will pitch in to pay for the hosting through Patreon, Open Collective, Ko-Fi, etc -- or mirror it themselves. Any streamer with a decent number of viewers will almost certainly have someone in the audience who is technically capable of running an instance if the streamer can't or doesn't want to DIY.
I get being on YouTube and Twitch -- PeerTube's discoverability sucks -- but for goodness sake, take ownership of your archives! If you make videos, that is your long tail! That is your legacy! Own it!
I don't use Twitch deeply so I'm not sure if this is a big enough thing to make people switch in large numbers, but if it is, the tech stack just doesn't seem like a moat at all these days. If anything, I'd say the fact that they prune old content already sort of is the opposite. YouTube's deep content library makes it hard to compete with them. Twitch purposefully doesn't even have one.
A skilled programmer could probably bang out a viable competitor in a week, and raise funds just as fast if the AWS bill became significant.
And the culture. Your examples are from the 2000s. The culture of the Internet back then was vastly different than it is today.
> A skilled programmer could probably bang out a viable competitor in a week, and raise funds just as fast if the AWS bill became significant.
I disagree. Where is this magic funds button? You're gonna need quite the pitch to get an investor to invest LOTS of money going up against Amazon (edit: and Google!).
Running a massive video site is not as simple as throwing a bunch of skilled programmers at it...
I'd guess it's something like 99% of content is seldom, if ever viewed, but I have no clue.
As for videos over 100 hours, it may be mostly top streamers.
So I think the reaction is because there's no now way to keep over 100 hours of video long term on Twitch?
If there's value in the VODs for content creators charge them for storage to at least break even, for VODs that don't get any views creators will have an incentive to delete them if they have to pay, problem solved. There's no need for arbitrary 100 hour limits or only targeting x% of creators, just use good old price signals.
I wonder why it's showing "2 hours ago" and it's now showing up.
And this submission shows up there (it's on the second page at the time that I write this comment)