mirror of
https://github.com/imputnet/cobalt.git
synced 2026-05-22 22:39:42 -05:00
[GH-ISSUE #1375] Can't download anything from YouTube on the main instance, again #12203
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @Spectaman on GitHub (Jun 30, 2025).
Original GitHub issue: https://github.com/imputnet/cobalt/issues/1375
bug description
When I tried downloading from a youtube link, I received this error:
couldn't get this video because youtube asked the processing instance to prove that it's not a bot. try again in a few seconds, but if it still doesn't work, please report this issue!I give it minutes before i try again but I get the same issue.
If it matters, the link I used was from YouTube Music specifically. I tried using the equivalent URL from regular youtube, but that didn't work either.
This never happened to me before. I didn't make any changes to my web browser as far as I know.
reproduction steps
screenshots
links
platform information
Librewolf on Linux Mint 22
additional context
No response
@BK-2-0-1 commented on GitHub (Jun 30, 2025):
same problem sadly. i love cobalt so much but it's not working on youtube for a couple weeks for me T_T
@Zoidborg7 commented on GitHub (Jul 1, 2025):
I am also getting this message every time I attempt any YouTube video. I have tried a few different videos, and I have waited between attempts a few times. I am reporting on this issue as it says in the instructions. Thanks!
Additional information: Windows 11 Pro OS. Edge browser (Chromium based; latest beta), also Google Chrome browser (latest non-beta)
@caohuan99 commented on GitHub (Jul 1, 2025):
tried using many IPs,same error
@VisionTQ commented on GitHub (Jul 1, 2025):
I have the same exact issue
@StrangerGX commented on GitHub (Jul 1, 2025):
At first everything worked perfectly, but just recently everything i tried has not worked
@wukko commented on GitHub (Jul 1, 2025):
just disabled the youtube support (again) until we figure something out (again), you'll see this error in the meantime (again):
@Intercity225 commented on GitHub (Jul 1, 2025):
i'm still getting hung up on cloudflare thinking i'm a bot
@Scriditer commented on GitHub (Jul 1, 2025):
Is there any good alternatives to cobalt, as i haven't used any others in a while.
@wukko commented on GitHub (Jul 1, 2025):
https://github.com/yt-dlp/yt-dlp
@MrCoreyTys commented on GitHub (Jul 1, 2025):
Well, I've been using this one called Open Video Downloader/youtube-dl-gui. Basically, it uses the code of youtube-dl and makes it so that you can download videos from wherever! like X, vimeo, and, of course, YouTube. It's really good for people who don't want to do command stuff because to some people, that can be a bit of malarkey. You can find it here: https://github.com/jely2002/youtube-dl-gui
@nrys-chrl commented on GitHub (Jul 1, 2025):
How come it only seems to be cobalt struggling with this when the other tools remain unaffected? genuinely curious
@gablilli commented on GitHub (Jul 1, 2025):
That gui has often some serious problems and it's not maintained anymore. Use something like Parabolic or the maintained fork of the original repo.
@slittle95 commented on GitHub (Jul 1, 2025):
Any tool I've seen that does not require you to install a program is broken: yt1s, y2mate, and cobalt have been my go-tos over the years, and none of them function right now.
@soulware1 commented on GitHub (Jul 2, 2025):
youtube making sure to mess up downloading every time i actually want to download something
@ledsales commented on GitHub (Jul 3, 2025):
As mentioned in the related thread, I'm using JDownloader 2, seems to be the best option at the moment, at least for me. Takes a bit of setting up, but once done, you can just throw links at it in bulk and it will grab the vids en-masse.
@TedKaczynskisCousin commented on GitHub (Jul 3, 2025):
I've seen another guy mention https://notube.link/ as a temporary alternative for now. Its just a tad bit sketchier than the sites you mentioned, but as long as you are using an adblocker and can look past (or translate) the foreign language, you should be pretty fine.
@gablilli commented on GitHub (Jul 3, 2025):
I even found an online gui of yt-dlp, but it isn't opensource and stores for an hour the downloaded files in their download folder on their server, but it actually works very well.
@slittle95 commented on GitHub (Jul 3, 2025):
@TedKaczynskisCousin notube is working nice and clean, with ublock no popups even, thanks for the rec! Actually I'd say this is less sketchy, haha, especially compared to y1ts which now outsources to a tiktok downloader under the hood--or tries to, the download won't complete until I open the tiktok downloader, and then the video it's downloading is apparently pulled from cache as it's always the first video I tried on it.
@gablilli gonna provide a link or nah?
@gablilli commented on GitHub (Jul 3, 2025):
Idk if I'm allowed to do this here.. I was actually the guy who mentioned notube, but my comment got deleted.
Anyways I'll try, here's the link.
@gablilli commented on GitHub (Jul 3, 2025):
just use the dot net version
@Tintin232 commented on GitHub (Jul 4, 2025):
@gablilli I tried that site, but I'm getting this error message when I hit "download file"
@slittle95 commented on GitHub (Jul 4, 2025):
@Tintin232 I got that a few times, but each time when I went back and reran the script it worked. I think it's not that it hosts each file for an hour, it's that every hour it cleans out its cache, and this happens when you get caught in the cleaning before you can download your file.
notube.link is faster by a long shot and doesn't do that, though it does semi-randomly complain at me about copyright. It does have a download limit, but it's quite high and you only have to wait two hours for it to reset. I've had both sites open for a couple hours working my way through a playlist.
@Tintin232 commented on GitHub (Jul 4, 2025):
@slittle95 I see, thanks.
@VisionTQ commented on GitHub (Jul 4, 2025):
Since Cobalt.tool isn't working, here are some alternatives I personally use:
If you're on Android, try ⭐seal. (APK)
On PC, use yt-dlp
If you're on iPhone... I honestly don’t know. Somebody tell me.
Someone mentioned:
⚠️ I haven't tested these links except seal & yt-dlp
@TheOnlyOnlyName commented on GitHub (Jul 5, 2025):
Hm I wonder how other sites still work
@LolproDoesStuff commented on GitHub (Jul 5, 2025):
I found an alternative website some days ago, it's literally cobalt but it runs on a community instance, it works perfectly and youtube still works.
you can try it out here: https://cobalt.meowing.de/
@lei-180 commented on GitHub (Jul 5, 2025):
@LolproDoesStuff trash website
it can process video, but whenever i tried to download, it's always err_invalid_handle
@TheOnlyOnlyName commented on GitHub (Jul 5, 2025):
Works for me, thanks!
@LolproDoesStuff commented on GitHub (Jul 5, 2025):
Sorry to say but I think it's a problem on your end.
I downloaded like 80 songs using that website and worked flawlessly so I think there is a problem with your machine or whatever.
@gablilli commented on GitHub (Jul 5, 2025):
same, it works very well.
anyways, any update on this? will this become only a tool that can be selfhosted (or deployed on some instances)?
@wukko commented on GitHub (Jul 6, 2025):
the main instance still supports 20 other services that work just fine. there are no news on youtube yet, but i’ll let you know when that changes!
@home-gihub commented on GitHub (Jul 6, 2025):
so the solution is to self-host cobalt to download from youtube?
@ThatRandomToast commented on GitHub (Jul 6, 2025):
Does this use the same method as IDM for downloading videos?
@VillainsRule commented on GitHub (Jul 6, 2025):
respectfully im following this thread for updates so if we could stop asking unrelated questions here that'd be great
@ThatRandomToast commented on GitHub (Jul 6, 2025):
This question is still a bit related, as IDM still download YouTube videos but cobalt.tools can't.
@AFasterSlowpoke commented on GitHub (Jul 9, 2025):
While cobalt fixes this, I found that FreeTube ( https://freetubeapp.io/ ) has an option to download mp4 or mp3 of youtube videos and it still works. it works on windows and linux mint. however, it can only download videos (with the mp3) in 360p, unless you download the video and audio seperately and merge them together.
Edit: some other guy commented about https://cobalt.meowing.de/ (which is literally just cobalt but a community hosted instance)
@jekeam commented on GitHub (Jul 10, 2025):
+1, What's the problem with linking personal local authorization data to YouTube? There are services like Pipe that have solved this problem a long time ago?
@MrCoreyTys commented on GitHub (Jul 12, 2025):
So, how's the progress currently with fixing the problem?
@practicenotes007 commented on GitHub (Jul 13, 2025):
Is there any new comment on the process currently with fixing the problem?
@SpiralNuggets commented on GitHub (Jul 13, 2025):
I'm sure that this is also a problem on self-hosted instances too. I have used the main frontend with custom processing and it shows the same error message. Direct API call into the instance returns
error.api.youtube.loginerror.@KwiatekMiki commented on GitHub (Jul 14, 2025):
only on some, e.g. it works on mine
@home-gihub commented on GitHub (Jul 14, 2025):
last time i checked https://cobalt.meowing.de/ works
@Ascellayn commented on GitHub (Jul 14, 2025):
Can confirm, yt session generator didn't help, I had to create a sacrificial google account and export the cookies and add said cookies to my cobalt instance for youtube to work.
@gablilli commented on GitHub (Jul 15, 2025):
same as cobalt.canine.tools.
Are there any updates for the main instance?
@XDuskAshes commented on GitHub (Jul 15, 2025):
Tried https://cobalt.canine.tools/ and it worked fine for video and audio download, I checked their instance info and found:
(
about-->instance information)@gablilli commented on GitHub (Jul 15, 2025):
oh didn't know that. I used that instance for downloading 50+ videos and it worked fine.
but even @KwiatekMiki 's instance works fine, so idk.
Let's wait for new updates.
@Tintin232 commented on GitHub (Jul 15, 2025):
I'm struggling to understand, how is it possible that different instances of cobalt work when the root of it all itself does not? Just wondering.
@XDuskAshes commented on GitHub (Jul 15, 2025):
The instance in particular (https://cobalt.canine.tools/) states (as quoted) that they "use a session server for youtube", so offloading YouTube things to a separate server. Since cobalt is FOSS it's reasonable to assume they made a tweak or two to make it work that way.
@Tintin232 commented on GitHub (Jul 15, 2025):
So why doesn't the main instance do it that way as well?
@Ascellayn commented on GitHub (Jul 15, 2025):
are you guys sure you aren't confusing said session server with https://github.com/imputnet/yt-session-generator which only aids but will still eventually lead to problems if you're using it too much
@ethrydevelops commented on GitHub (Jul 16, 2025):
said this in the other (dead) thread, it might be useful to implement idk
@squabbled commented on GitHub (Jul 17, 2025):
It is a good idea in concept, but the overhead of using TOR would almost certainly make it less effective. I tested YouTube in TOR and the results were fairly inconsistent, some nodes would not prompt a sign in but others would, which is similar behavior to Mullvad, some servers do require sign in while some don't, usually dependent on server traffic. The TOR network is good for anonymity but is typically fairly slow compared to a direct connection or VPN. Right now testing without any VPN I get 120mbps down while with TOR I get 14mbps down. TOR's also used by a lot of activists and users in restricted countries so running a large downloading service could also cause this speed to deteriorate for others too. There's a reason that large data actions like torrenting is heavily frowned upon among TOR users.
@home-gihub commented on GitHub (Jul 18, 2025):
how about downloading from invidious (not thru)?
@sprague60 commented on GitHub (Jul 19, 2025):
Alas, i guess its time to move on to other services. As much as i like cobalt.... can't really hold my breath that much longer for a workaround. c'est la vie. nothing personal.
@gablilli commented on GitHub (Jul 19, 2025):
Nah, I have to disagree. After more than 8 years of downloading videos from YouTube, I haven’t found a better solution than Cobalt.
It’s safe, clean, open-source, fast, has a great UI, and is cleverly built. Right now, I’m using some of the Cobalt community instances, which work just as well.
I’m not sure what alternative you found that made you say it’s time to move on, but from what I’ve seen, there really aren’t any other online services this good without being sketchy.
Here's my take — I still think Cobalt (or its forks) are well worth sticking with. Just wait for an update if you want the main instance working again, and in the meantime, use the community instances (or maybe set up one of your own).
@gablilli commented on GitHub (Jul 19, 2025):
I totally get the frustration, but honestly, I think the situation is more straightforward (and less “sketchy”) than it might seem.
The reason of all of this could be found here, in the "YouTube downloading" section, or even here!.
That said, it would be nice to hear an update or even just a brief explanation from @wukko about what’s going on — especially since the community forks have been working for weeks. Even just a bit of insight into the current limitations, an ETA, a "general" update on the situation would go a long way.
Now, I’ll be the first to say: in extreme cases, I do suggest using alternative sites if nothing else works. I get it — sometimes you just need that one download quickly. But whenever I know the address of a working Cobalt instance, I’ll always choose that over anything else. Tools like NoTube or SaveFrom should honestly be a last resort in this context.
From my experience, those platforms are often filled with pop-ups, shady ads, and in worst-case scenarios, they can be genuinely unsafe. I remember a Reddit post where someone said they found a malicious PowerShell script copied into their clipboard after downloading from NoTube. Might’ve been a one-off, but still — not a great look.
To be fair, NoTube has worked well in the past and still gets the job done. But the fact remains: it’s not open-source, and we really don’t know what’s going on behind the scenes — what data is tracked, how downloads are processed, etc. Cobalt, on the other hand, is transparent, clean, and well-maintained — and for people who care about that, it makes a big difference.
Also worth noting: a lot of those commercial download sites change domains constantly, and that’s usually not a great sign. Legal pressure, security issues, shady monetization — who knows. And realistically, if YouTube is tightening restrictions now, those other sites will probably get hit eventually too.
So yeah, I’m not saying everyone has to use Cobalt — people should use whatever works best for them. But from where I’m standing, I still haven’t found anything else that matches its balance of safety, speed, and transparency. Just my two cents.
@VillainsRule commented on GitHub (Jul 19, 2025):
one again, i am requesting you all take the fighting to elsewhere
many people are following this thread for updates and not people fighting over stupid stuff
@squabbled commented on GitHub (Jul 19, 2025):
Invidious also faces many of the same issues as Cobalt when it comes to YouTube access, unfortunately. Many Invidious instances get the same "Sign in to protect our community" error as Cobalt and typically have to use residential proxies (which are expensive, and even those are often blocked too). If Cobalt were to host an internal Invidious instance(s) and scrape that, it would likely get the same errors at the same rate that it gets them with accessing YouTube directly.
@firsturdiev commented on GitHub (Jul 21, 2025):
Even cobalt contributors are able to find a reliable solution for YouTube downloads in the future, I'm strongly against open-sourcing it, better to keep that solution code private.
Because you know there are thousands of website owners who scrape Youtube to make money by ads. They constantly watch for any open source projects for finding a solution, they don't even donate a single dollar but use the code in their way.
I had a conversation with a lot of them, large portion of them even don't know programing, they just hire one-time freelancers for new updates that's why their website & service terrible (some people already mentioned such websites at the top). The reason why they're still able to download videos in such tight scenarios is that they have a huge amount of income and they're ready to spend even half of it for just keeping the service alive & not affecting SEO.
If cobalt make Youtube downloading code private, such websites are left with no food, and people see cobalt.tools instead of ad-ful & scam websites in the search results. Plus this approach improves stability of that solution because it'll be used by only cobalt.tools, not many scrapers, so YouTube might not lift up barriers for that solution.
It's just personal opinion, not disrespecting or forcing anyone. What's your opinion on my thoughts?
@ethrydevelops commented on GitHub (Jul 22, 2025):
@firsturdiev it seems like they're targeting cobalt in particular, so youtube can always scan logs to see how cobalt is bypassing their filters
@firsturdiev commented on GitHub (Jul 22, 2025):
Youtube staff actively monitors code & GitHub issues of yt-dlp, it's once said in one of issue discussions. I don't get surprised if they also monitor Cobalt's.
That's the point why I'm saying disclosing the download method in order to keep alive cobalt.tools.
@bashonly commented on GitHub (Jul 22, 2025):
YT is blocking IP addresses that are suspected of automated downloading/scraping, and the blocking is trending more and more aggressive. It's not any deeper than that. There's no magic solution to this problem that would require secrecy.
@firsturdiev commented on GitHub (Jul 22, 2025):
I launched my instance exclusively for YouTube, check it, appreciate your feedbacks: https://pictube.app
@p0008874 commented on GitHub (Jul 29, 2025):
https://github.com/user-attachments/assets/f43c60f2-ce2e-4070-961a-01edac98293f
Not sure if this can help you, you can spoof the user agent by using Firefox, then switch back to normal.
@wukko commented on GitHub (Aug 15, 2025):
all future talk about this issue should be in #1356, since it's the original issue about the underlying problem. strict network limits on youtube's side are still present, and we can't go around them at our scale. we will have to come up with a completely different way of interacting with innertube that doesn't involve a centralized "proxy" server. using google accounts is not an option, they get restricted faster than we can replace them, and it's an insane amount of manual labor nearing slavery.
please don't spam that issue with random stuff not related to the topic, all complaints and nags will be hidden as off-topic.