Using audioonly filter slows down download tremendously
nikbrandt opened this issue · 12 comments
When downloading a video and converting to mp3 with ffmpeg, using filter: 'audioonly' in the options slows down speeds by magnitudes; ~72 seconds with filter, ~8 with no options, also ~8 with quality: 'highestaudio' (seems weird considering highestaudio requests just audio, like the filter).
This is tested using a 4:13 video on my gigabit ethernet.
code used:
const ytdl = require('ytdl-core');
const ffmpeg = require('fluent-ffmpeg');
let id = 'sDLsSQf3Hc0';
let start = Date.now();
let stream = ytdl(id, { quality: 'highestaudio'/*, filter: 'audioonly' */}); // mess around with quality, filter, and nothing at all being used
ffmpeg(stream)
.on('end', () => {console.log('done, thanks - ' + (Date.now() - start) / 1000 + 's')})
.audioBitrate(128)
.save(`${__dirname}/${id}-${(Math.random() * 100000).toFixed(0)}.mp3`); // very efficient naming system hereThis doesn't really affect me because I can just use quality: 'highestaudio', but it does seem like a weird effect.
i tried running your code it keeps giving me this issue
Could not extract signature deciphering actions
This is a bit strange, some formats download way faster than others. Even when they're bigger in filesize. From the following list, the top 3 downloaded super fast. Then, it significantly slows down after that, but not slow enough that it would be slower than it plays.
itag container resolution video enc audio bitrate audio enc
22 mp4 720p H.264 192 aac
43 webm 360p VP8 128 vorbis
18 mp4 360p H.264 96 aac
36 3gp 240p MPEG-4 Visual 32 aac
17 3gp 144p MPEG-4 Visual 24 aac
299 mp4 1080p H.264
303 webm 1080p HFR VP9
137 mp4 1080p H.264
248 webm 1080p VP9
298 mp4 720p H.264
302 webm 720p HFR VP9
136 mp4 720p H.264
247 webm 720p VP9
135 mp4 480p H.264
244 webm 480p VP9
134 mp4 360p H.264
243 webm 360p VP9
133 mp4 240p H.264
242 webm 240p VP9
160 mp4 144p H.264
278 webm 144p 15fps VP9
251 webm 160 opus
140 m4a 128 aac
171 webm 128 vorbis
250 webm 64 opus
249 webm 48 opus
And I'm not sure there's much we can do about it. Since it still downloads faster than one can play a video, even on 2x, this rate limit could be intended. Although I'm not sure why it only happens for some formats.
There is the ratebypass param. There might be more params like it.
Also, thank you for this code snippet. I'd say it's pretty common, and so I'll add it to the examples folder.
Alright, cool.
How would you find out about more parameters? Shuffle through all the info in .getInfo for the video? I'm willing to do that, lol.
@bolahanna44, I'm pretty sure you just need to update ytdl. (npm update ytdl-core or npm i ytdl-core@latest)
This might be obvious from the earlier testing you did, but looking at the formats;
those first three formats had ratebypass in the URL in two places.
They had it after the itag part of the link (... ?itag=18&ratebypass=yes ...) and towards the end of the deadly URL, before the requiressl part (... %2Cratebypass%2Crequiressl ...).
When I added ratebypass to those parts of a URL that did not have 'em, I was met with a lovely 403 Forbidden page.
Probably is intended that only those formats allow ratebypass, it's really bizarre to wonder why though.
Did you also add it to sparams?
Sorry for such a delayed reply
adding it to sparams => 403
not adding it to sparams => since ratebypass=yes is already in the query we're not changing anything
Yes, I also added it to sparams - ^ what they said
changing, say, a working url
https://r1---sn-nvopjoxu-nune.googlevideo.com/videoplayback?lmt=1517944841813447&key=yt6&itag=251&keepalive=yes&fvip=1&c=WEB&requiressl=yes&ip=184.23.255.99&mv=m&mt=1528232751&ms=au%2Crdu&mn=sn-nvopjoxu-nune%2Csn-n4v7sn7z&mm=31%2C29&pl=18&mime=audio%2Fwebm&id=o-AK1via8p26fKTNcSBEio9T87CCTx9TZBDGjVHhGbLA3j&ipbits=0&expire=1528254485&dur=253.101&initcwndbps=750000&ei=tfsWW9b_EPXI-APw0ZeQCg&clen=4486749&source=youtube&gir=yes&sparams=clen%2Cdur%2Cei%2Cgir%2Cid%2Cinitcwndbps%2Cip%2Cipbits%2Citag%2Ckeepalive%2Clmt%2Cmime%2Cmm%2Cmn%2Cms%2Cmv%2Cpl%2Crequiressl%2Csource%2Cexpire&signature=4A31B3796440A7D517654DCDE8CAE00F682CAC51.367CB30186ADD857BBC9DE8B36C3BAB56E75C3F6
to include ratebypass as both a param and a sparam
https://r1---sn-nvopjoxu-nune.googlevideo.com/videoplayback?lmt=1517944841813447&key=yt6&itag=251&keepalive=yes&fvip=1&c=WEB&ratebypass=yes&requiressl=yes&ip=184.23.255.99&mv=m&mt=1528232751&ms=au%2Crdu&mn=sn-nvopjoxu-nune%2Csn-n4v7sn7z&mm=31%2C29&pl=18&mime=audio%2Fwebm&id=o-AK1via8p26fKTNcSBEio9T87CCTx9TZBDGjVHhGbLA3j&ipbits=0&expire=1528254485&dur=253.101&initcwndbps=750000&ei=tfsWW9b_EPXI-APw0ZeQCg&clen=4486749&source=youtube&gir=yes&sparams=clen%2Cdur%2Cei%2Cgir%2Cid%2Cinitcwndbps%2Cip%2Cipbits%2Citag%2Ckeepalive%2Clmt%2Cmime%2Cmm%2Cmn%2Cms%2Cmv%2Cpl%2Cratebypass%2Crequiressl%2Csource%2Cexpire&signature=4A31B3796440A7D517654DCDE8CAE00F682CAC51.367CB30186ADD857BBC9DE8B36C3BAB56E75C3F6
still leads you to the 403
I'm not entirely sure there's anything we can do about this; it seems this is in place for a reason ¯\(ツ)/¯
Yeah. I'll close this and add a note to the readme.
Hi a question, does anyone tested it with multi requests?
Hi, I suppose you investigated this, but youtube-dl doesn't seem to suffer from rate limiting. I think you can copy its behaviour in ytdl-core.
I looked into it a little bit more, and it seems that adding a Range header, e.g.
fetch(u, {headers: {Range: 'bytes=0-10380331'}})(here, using node-fetch) improves download speed dramatically.