Regarding Youtube-dl download speed is limited by the YouTube website #30097
Comments
|
I came here to look in to the same issue. grabbing 4k (so webm as well) content is heavily throttled. edit: disregard, read the other threads and issue is fixed using the yt-dlp fork/successor |
|
Duplicate of #29326. |
|
@dirkf |
Hi! One of many examples: I can fully confirm the issue. It is ongoing since (at least) several days). Most videos that I tried during that time come in with a maximum of 80 KB/s. Cheers! |
|
I'm having the same problem, for the past week or two I can't seem to get any speeds from youtube greater than 100-200 kb/s |
|
I too noticed slow downloads from youtube since a week or two |
|
same, slow downloads from youtube (80 kb/sec) |
|
I'm experiencing the same issue since a week or so. Previously I would sometimes have issue with choppy video when the video wasn't completely cached (sometimes just running desired quality and restarting it after a few minutes would fix the issue). Now it's unescapable. |
|
just use yt-dlp |
|
So the Android client hack still works despite the apparently stronger throttling and should be back-ported? |
No, apparently now it depends on the reporter. Maybe your IP has some history of regular large downloads? Maybe no-one, not even YT staff, knows why the throttling occurs because it depends on some deep learning algorithm? I suggest that implementing either the challenge-response algorithm in the linked thread or (for the moment?) using the Android client bypass from yt-dlp should resolve this issue. |
|
I tried different IP addresses and found that the speed limit was between 50 KB/s and 150 KB/s. |
|
I can confirm that this happens on our kubernetes nodes and local machines as well. The speed is limited to around 50KB/s. We operate under 1Gbps networks. |
|
|
I can confirm that yt-dlp works flawless with videos that get throttled via youtube-dl. But that does not solve the problem, I guess. However, nice workaround for the meantime! |
Actually it does. This project has had its last commit more than 3 months ago, a huge amount of time for a program that depends on external changing sites. On Linux you can basically symlink From Python code you can do: TLDR: It's dead Jim. Move to |
Ideally, we ofc want a way to solve the js challenge. For this, we either (1) hardcode the js challenge (this is a maintainability nightmare), (2) write a more complete js interpretter (there is already a PR here, but its quite old and no longer up to spec) or (3) use some external dependancy. But that said, the android workaround is not expected to break in the near future because google wont be able to implement the challenge in their android app without making the all prior versions of the app unusable. So any implementation of the challenge on android is expected to be a slow rollout and we can therefore expect to get a reasonable amount of notice before this starts becoming a problem |
Would be very useful to get that working given the limitations of the existing |
|
Just curious, how exactly did yt-dlp solve this issue? |
A quick summary: https://code.videolan.org/videolan/vlc/-/issues/26174#note_286445 |
|
Hello, I'm the VLC maintainer of the YouTube playback feature in the VLC media player. There's been talk about javascript interpreters, and I wanted to share how you don't need to run the descrambling code through a javascript interpreter to make it work. In VLC, extractors are website parser scripts written in Lua. We emulate the javascript descrambling code within the Lua script using ad hoc code, on the basis that it uses only a known set of structured transformations. This approach is less reliable and requires more maintenance than a generic interpreter as slight changes in the code or javascript minifier make it break every now and then; but it's worked for us for the past 8 years for signature descrambling, and we just released a version that descrambles the "n" parameter too that's required to solve this throttling. So this is a proven concept, and another possibility. If you want to take a look, see the https://code.videolan.org/videolan/vlc/-/blob/master/share/lua/playlist/youtube.lua Of course, as long as playback in VLC works, it's always another fallback alternative to youtube-dl to use VLC as the URL extractor, and then copy-paste the working direct video file URL from VLC to pass it to |
|
@linkfanel, I've made a WIP Python Sadly I must be missing something as the descramble is getting scrambled (eg, trying to rotate "n" by a non-integer value). This is a failing transformation log for n = Do you perhaps have a log that you could share so that I could do an A-B comparison? |
|
@dirkf pytube also has an implementation of the descrambling algorithm. Since it is in python too, it may be a better reference |
I bet. In Lua no table length built-in function, no table reverse built-in function, table indices starting at 1... Here's what I get: Does your rotate transformation handle 0 correctly? Your data array seems fine up to that point. |
|
Excellent, thanks. As elsewhere, there can be So ~500kB/s on this tiny platform on my LAN where unconstrained downloads rarely beat 1.5MB/s. @pukkandan said:
Although this is close to (1) as a solution, based on VLC's experience with the signature algorithm it may not be quite such a nightmare and the approach from the To close off the discussion here, this |
|
so should we wait for new release? |
Fixes ytdl-org#29326, closes ytdl-org#29790, closes ytdl-org#30004, closes ytdl-org#30024, closes ytdl-org#30052, closes ytdl-org#30088, closes ytdl-org#30097, closes ytdl-org#30102, closes ytdl-org#30109, closes ytdl-org#30119, closes ytdl-org#30125, closes ytdl-org#30128, closes ytdl-org#30162, closes ytdl-org#30173, closes ytdl-org#30186, closes ytdl-org#30192, closes ytdl-org#30221, closes ytdl-org#30239, closes ytdl-org#30539, closes ytdl-org#30552.
Recently, when using Youtube-dl, it was discovered that the download speed of some content was restricted. Generally, it is controlled at 80kbps/s-35kbps/s. I don't know how to fix this fault. In particular, the webm format video and audio content are faulty. Other formats have not been tested yet. How to solve it! Thanks!
The text was updated successfully, but these errors were encountered: