Going through the linked issue, it seems that the fix has already been merged and should be part of the next invidious release.
But it still is a fact that Google is trying to limit access to YouTube through alternatives like yt-dlp and invidious by constantly trying to break stuff.
It's long been known that Google's weapon is change; they've been doing that with web "standards" too. For YouTube, I guess the next stage in the cat-and-mouse game is to automate fixing it, and I even dare say that this new AI stuff might come in handy for that.
The culture I find myself in, in 2024, is one of war; unrecognisable
from anything I've seen in "peaceful civilian life" before. It has no
resemblance to business, to markets, to value, to a lawful society of
fairness, utility and cohesion. What happened to us?
I don’t think anything happened to us, so much as your memory is flawed. We’ve never evolved beyond base self-interest, we’ve just become much more productive at extracting value from every aspect of life.
I'm genuinely sorry for you that you feel your culture has become
nothing but an extractive free-for-all. There remain people who still
aspire to treating others as ends and not means. Extraction is not
productivity. To produce is to create. Your comment kinda validates
what I was hinting at more though; that BigTech now sees itself as in
all-out "do what though wilt" conflict with the civil population. It's
a cyber war in all but name - though we still pretend it's
"business". At some point the population are going to figure the rules
have changed. From that point I'd give FAANG a few months to a
year. Most analysts have figured that a cyber war would be
inter-national rather than intra/civil conflict. We'll need a backup
plan for Google/Microsoft etc unless they're radically neutralised (by
government breakup or similar)
My hypothesis is that Invidious is small, unimportant and unprofitable enough that Google just does nothing to care for it. It's not being tested, there's no QA, and the impact on Invidious isn't considered at all when planning API changes.
If you're taking a walk in the forest, you don't care if you're crushing insects under your feet. You do nothing to deliberately crush them, but you don't walk extremely slowly all the time in case there's an insect nearby.
How are you using it? If you come across a youtube link, do you run yt-dlp on the command line and then watch the downloaded file in a media player? Asking because I'm too lazy for that, and interested to know if there is some fancier way.
Setting a lower resolution format via "ytdl-format=..." in the mpv.conf file may resolve that. Personally I also use "pause=yes" and find that the few extra seconds it has to cache also helps.
I usually like to promote cool FOSS that I use, and Newpipe is a masterpiece, but I do feel reluctant to increase the popularity of something that relies on unpopularity to survive.
Yes, exactly that. So much stuff just vanishes, it's prudent to simply download everything by default. Whether that's a video with yt-dlp, or a webpage with singlefile or other tools, save it all.
I wish I had better tools for simply saving everything I see on the internet at all times. The disk-space cost would be trivial in the era of terabyte drives, and simply never worrying about link-rot would be huge.
I found grab-site to be extremely effective with even complex, content-heavy websites. Dunno about YouTube though, I’m afraid of what would happen if I tried running such an archival routine. My server would probably catch fire.
What I really want is a thing that goes through my browser history and just archives everything unless I specifically exclude it.
Browser history _used_ to do that, almost. You could browse your history offline and pages would still render. And at some point, that stopped working.
It turns out that it's easy to rig eww (Emacs's built-in browser) to automatically pass URLs to whatever special handlers you desire, via the "browse-url-handlers" variable. I did that earlier this weekend for Github and Youtube. For Youtube, I have a wrapper function just starts mpv with the supplied url. For Github, a different wrapper just breaks out the user/repo pair from the url and invokes github-explorer, available in MELPA. (When I need more from a repo, I can just clone it and explore it locally. Either way, there's little need to deal with Github's web UI.)
Not quite. Late last year I began running a Searx instance locally, less for privacy than for having a hackable meta-search engine. Since it supports several "bangs," like !ddg and !yt and the like, I generally use this for quickly searching Youtube without changing context. When I enable images [0], I can see the thumbnails just fine. Regardless, there are plenty of places on the web (like here) where people will post interesting YT links, and having them pop open in mpv is a good default [1].
I've also evolved a CLI YT client, based on mpv/yt-dlp and fzf; it does not show thumbnails. Even so, it's perfectly fine for gathering lots of search results, like
$ yt 500 medical uses for hollandaise
which will gather the first 500 results about medical hollandaise [1], filter out the shorts, and present the remaining results in an fzf-based UI. I haven't really missed the thumbnails.
---
[0] One of the first things I did with eww was to bind "I" to a function that toggles images on or off and re-displays the current page.
[1] In my setup, I can override all special handling of a URL by holding Ctrl while hitting Enter to follow a link.
[2] Earlier today, I had a fascinating conversation with TinyLLama about the medical uses of hollandaise and options for regulating it. Did you know that it can be taken intravenously to help fight cancer?
Been running a non-public Piped instance for a couple of years now.
There are videos where it works and videos where it doesn't.
It's been a few months now where some proportion of videos fail to play over Piped no matter what (the same videos will fail across public instances too) and anecdotally that percentage has increased recently.
> It is likely that existing alternatives and new alternatives will see huge user jumps when that happens.
I've heard that often and i'm pessimistic about it in general. Although there might be some hope, because the channels Invidious users might prefer (nerd stuff) are more likely to move to an alternative, than others.
Google really needs to revamp their YouTube model. I'd sign up for an embedding API with a low user fee (~$0.50) and generous low volume (free) and reasonable high volume (>50k requests/month at $0.01/1k). Ad free, of course, because YouTube ads are probably not great money makers for small businesses, hence the awful commercials we see pushed and successful creators reliance on in-video promotions.
They are artificially suppressing the api value by banning uses people want. If people want to opt out of the recommendation engine and ads that makes hosting cheaper and easier for YouTube. They only do it to leverage market position and imo that's bad.
Perhaps you didn't catch my comment's assertions. I stated the model and approximate values I'd be willing pay, given that most creators take in in-stream paid promotions already (above and beyond YT ads) and that general YouTube ads have poor power/are garbage/have low signal to noise ratios.
I don't really care what Google attributes to YouTube indirect or direct revenue and costs on their P&L in the context of average revenue, costs, and approximate minutes streamed, since the marginal costs are ridiculously cheap.
There is a lot of value in Google adopting the first mover advantage in a Bernoulli market. I'd imagine shareholders care about that a lot.
> Perhaps you didn't catch my comment's assertions. I stated the model and approximate values I'd be willing pay
No I understood it perfectly, it is just absurd. You lack power, leverage, or influence over Google so the only lever you have to make a better deal is by offering more money than you are currently to get the terms you'd like.
I don't need power, leverage, or influence to state level of willingness to pay. That's independent of whether Google chooses to service the market segment I belong to. I'm part of the demand side, not the supply side.
They should still rethink their model, regardless.
Could you point to the source on that figure? I gather streamed means simply “showed to an individual” in this context, but is this figure official or an estimate and does it take jnto account actual viewing time or is it based on an estimate of virwijg time and total video views?
A parameter was changed from '2AMBCgIQBg' to 'CgIIAdgDAQ%3D%3D' which is just the correct base64 encoding they should have been using the entire time.
I don't think this was a hostile action by Google. An engineer probably just fixed a bug with input validation which broke the malformed requests Invidious was sending.
I wonder how long it took G to push out this breaking change. I wonder if there will ever be a point where G will just give up.
The fight between G and alternative front ends reminds me of the over engineered solution by NYC MTA to prevent people from hopping the turnstile.
They invested some millions of dollars into building and deploying the system. Only for the system to be a very mild deterrent which can be easily fooled.
67 comments
[ 3.4 ms ] story [ 147 ms ] threadBut it still is a fact that Google is trying to limit access to YouTube through alternatives like yt-dlp and invidious by constantly trying to break stuff.
It's long been known that Google's weapon is change; they've been doing that with web "standards" too. For YouTube, I guess the next stage in the cat-and-mouse game is to automate fixing it, and I even dare say that this new AI stuff might come in handy for that.
> > Google's weapon
The culture I find myself in, in 2024, is one of war; unrecognisable from anything I've seen in "peaceful civilian life" before. It has no resemblance to business, to markets, to value, to a lawful society of fairness, utility and cohesion. What happened to us?
I'm genuinely sorry for you that you feel your culture has become nothing but an extractive free-for-all. There remain people who still aspire to treating others as ends and not means. Extraction is not productivity. To produce is to create. Your comment kinda validates what I was hinting at more though; that BigTech now sees itself as in all-out "do what though wilt" conflict with the civil population. It's a cyber war in all but name - though we still pretend it's "business". At some point the population are going to figure the rules have changed. From that point I'd give FAANG a few months to a year. Most analysts have figured that a cyber war would be inter-national rather than intra/civil conflict. We'll need a backup plan for Google/Microsoft etc unless they're radically neutralised (by government breakup or similar)
If you're taking a walk in the forest, you don't care if you're crushing insects under your feet. You do nothing to deliberately crush them, but you don't walk extremely slowly all the time in case there's an insect nearby.
> mpv with a yt-dlp hook
What does hook mean in this context? Is it some kind of plugin or addon to MPV or can MPV do that out-of-the-box?
1. A desktop application call yt-dlg/youtube-dl-gui.
2. Metube, which is a selfhosted application and there is a plugin for the browser to send things to it.
I wish I had better tools for simply saving everything I see on the internet at all times. The disk-space cost would be trivial in the era of terabyte drives, and simply never worrying about link-rot would be huge.
Browser history _used_ to do that, almost. You could browse your history offline and pages would still render. And at some point, that stopped working.
I've also evolved a CLI YT client, based on mpv/yt-dlp and fzf; it does not show thumbnails. Even so, it's perfectly fine for gathering lots of search results, like
which will gather the first 500 results about medical hollandaise [1], filter out the shorts, and present the remaining results in an fzf-based UI. I haven't really missed the thumbnails.---
[0] One of the first things I did with eww was to bind "I" to a function that toggles images on or off and re-displays the current page.
[1] In my setup, I can override all special handling of a URL by holding Ctrl while hitting Enter to follow a link.
[2] Earlier today, I had a fascinating conversation with TinyLLama about the medical uses of hollandaise and options for regulating it. Did you know that it can be taken intravenously to help fight cancer?
There are videos where it works and videos where it doesn't.
It's been a few months now where some proportion of videos fail to play over Piped no matter what (the same videos will fail across public instances too) and anecdotally that percentage has increased recently.
> It is likely that existing alternatives and new alternatives will see huge user jumps when that happens.
I've heard that often and i'm pessimistic about it in general. Although there might be some hope, because the channels Invidious users might prefer (nerd stuff) are more likely to move to an alternative, than others.
If you'd be willing to pay more than that to watch videos I am sure they'd take your money.
Wolfram Alpha helpfully formatted it as "0.14" and I missed the cent sign. https://www.wolframalpha.com/input?i=%2431.5+billion+%2F+21%...
I don't really care what Google attributes to YouTube indirect or direct revenue and costs on their P&L in the context of average revenue, costs, and approximate minutes streamed, since the marginal costs are ridiculously cheap.
There is a lot of value in Google adopting the first mover advantage in a Bernoulli market. I'd imagine shareholders care about that a lot.
No I understood it perfectly, it is just absurd. You lack power, leverage, or influence over Google so the only lever you have to make a better deal is by offering more money than you are currently to get the terms you'd like.
They should still rethink their model, regardless.
0.0014 cents per minute is $0.000014 per minute.
I don't think this was a hostile action by Google. An engineer probably just fixed a bug with input validation which broke the malformed requests Invidious was sending.
https://gchq.github.io/CyberChef/#recipe=URL_Decode()From_Ba...
The fight between G and alternative front ends reminds me of the over engineered solution by NYC MTA to prevent people from hopping the turnstile.
They invested some millions of dollars into building and deploying the system. Only for the system to be a very mild deterrent which can be easily fooled.
More discussion: https://news.ycombinator.com/item?id=39882455
This is just google being google and changing stuff.