m3u8 stream to mp4 using ffmpeg

m3u8-to-mp4.md

1. Copy m3u8 link

2. Run command

echo "Enter m3u8 link:";read link;echo "Enter output filename:";read filename;ffmpeg -i "$link" -bsf:a aac_adtstoasc -vcodec copy -c copy -crf 50 $filename.mp4

Example final command

ffmpeg -i "http://host/folder/file.m3u8" -bsf:a aac_adtstoasc -vcodec copy -c copy -crf 50 file.mp4

@magicdawn, most of the ffmpeg options here don't do anything so it's particularly confusing trying to understand them.

In most all situations you just need this:

ffmpeg -i "http://host/folder/file.m3u8" -c copy file.mp4

or

echo "Enter m3u8 link:";read link;echo "Enter output filename:";read filename;ffmpeg -i "$link"  -c copy $filename.mp4

to breakdown the original command...

-c copy -vcodec copy

is redundant. -c copy says "for all streams, do a stream copy" (no re-encoding). And then -vcodec copy says "for the video stream[s], do a stream copy". -c copy is all that's needed. You could also do -vcodec copy -acodec copy or -codec copy since -c is just a short version of -codec.

-crf 50

As you pointed out, since there's no transcoding going on, this video quality flag does nothing. And if it did work, 50 is such low quality that you probably wouldn't like the result. Also this flag is only supported by some codecs. (I wish ffmpeg gave you a warning when you pass in a useless flag!)

-bsf:a aac_adtstoasc

These days ffmpeg is smart enough to automatically insert this filter when needed. IMHO it's cleaner to just rely on that rather than always inserting it even when it's not necessary. (like in non-AAC+MPEG-TS streams) If you turn on verbose messaging with -v verbose you'll see that ffmpeg informs you of this:

$  ffmpeg -i https://...m3u8 -v verbose -c copy test.mp4
...
Automatically inserted bitstream filter 'aac_adtstoasc'; args=''

You can pull all this together and test things by downloading a stream with and without all these options and see that the resulting files are identical!

$ ffmpeg -i https://../stream.m3u8 -c copy test1.mp4
$ ffmpeg -i https://../stream.m3u8 -bsf:a aac_adtstoasc -vcodec copy -c copy -crf 50 test2.mp4
$ md5 test*.mp4
MD5 (test1.mp4) = 854e2beca4b3e516dd4cf147ca0521ee
MD5 (test2.mp4) = 854e2beca4b3e516dd4cf147ca0521ee

---

Another tip: if youtube-dl works but you want to control how ffmpeg downloads a stream without having to manually snoop for the .m3u8 file, just use youtube-dl's -g flag and it'll print out the m3u8 url and exit.

$ youtube-dl -g "https://host.com/some-webplay-video"

for example

$ youtube-dl -g https://www.pbs.org/video/east-bay-hip-hop-dance-in-richmond-ca-ng3v7e/
https://ga.video.cdn.pbs.org/videos/if-cities-could-dance/a7c8f330-2de4-4476-bb64-b23ae0bb2951/2000112404/hd-16x9-mezzanine-1080p/3f1sh5t6_richmond_iccd-16x9-1080p-1080p-6500k.m3u8

$ ffmpeg -i https://ga.video.cdn.pbs.org/videos/if-cities-could-dance/a7c8f330-2de4-4476-bb64-b23ae0bb2951/2000112404/hd-16x9-mezzanine-1080p/3f1sh5t6_richmond_iccd-16x9-1080p-1080p-6500k.m3u8 -c copy pbs-vid.mp4

When I use this command:
ffmpeg -i "http://host/folder/file.m3u8" -c copy file.mp4

The resulting file.mp4 is always a video of different length -- sometimes 20min, sometimes, 15min, sometimes 5. Not only this, but the videos do not appear to start at the same time. Anyone know what could be happening?

$ youtube-dl -g https://www.pbs.org/video/east-bay-hip-hop-dance-in-richmond-ca-ng3v7e/
https://ga.video.cdn.pbs.org/videos/if-cities-could-dance/a7c8f330-2de4-4476-bb64-b23ae0bb2951/2000112404/hd-16x9-mezzanine-1080p/3f1sh5t6_richmond_iccd-16x9-1080p-1080p-6500k.m3u8

$ ffmpeg -i https://ga.video.cdn.pbs.org/videos/if-cities-could-dance/a7c8f330-2de4-4476-bb64-b23ae0bb2951/2000112404/hd-16x9-mezzanine-1080p/3f1sh5t6_richmond_iccd-16x9-1080p-1080p-6500k.m3u8 -c copy pbs-vid.mp4

This is exactly what youtube-dl does if you just invoke

$ youtube-dl https://www.pbs.org/video/east-bay-hip-hop-dance-in-richmond-ca-ng3v7e/

i.e. it uses ffmpeg (or avconv if you so wish) in the background. No need to do that manually.

i got Output file #0 does not contain any stream error, not sure why

I received this response from ffmpeg

Then I tried: youtube-dl --no-part --restrict-filenames --user-agent "Mozilla/5.0 (Windows NT 6.1; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/78.0.3904.108 Safari/537.36" "https://XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX/playlist.m3u8?pulse=assetsXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX"
And it worked, then I tested the original comand line (ffmpeg) and now it worked without problems
Don't know why. 🤷‍♀️

Also, the map option could be used to select specific program (group of streams / quality) from the m3u8 file: -map p:2.

Tried using VLC and it was being tempermental on me, this worked so much better thank you

VLC interrupts the recording sometimes, but much more simple.

@varenc thanks for this break-down. Wouldn't ffplay "$link" essentially do the same thing? (No need to use ffmpeg)

How do I download m3U8 files that are encrypted using key? Such as:

#EXT-X-KEY:METHOD=AES-128,URI="https://xxxxx.com/somepath/somekey.key",IV=0x00000000000000000000000000000000

@varenc thanks for this break-down. Wouldn't ffplay "$link" essentially do the same thing? (No need to use ffmpeg)

If you just want play the stream and watch it once, I agree it is! But then you might as well use mpv or VLC to play the stream. I figure people finding this gist want to download the stream and keep the media around for later viewing too.

Thanks for the snippet, saved my day 👍

ffmpeg newbie here but, can I ask why -crf 50 was used when the encoder is copy ? Will it make a difference?

ffmpeg newbie here but, can I ask why -crf 50 was used when the encoder is copy ? Will it make a difference?

You're right, it does nothing! Since there's no encoding going on and the stream is just being copied, -crf 50 has no effect. I broke this down more in my comment above.

This command is all you need: ffmpeg -i "http://host/folder/file.m3u8" -c copy file.mp4

This command is all you need: ffmpeg -i "http://host/folder/file.m3u8" -c copy file.mp4

yes, but it only picked up the English audio in my stream just now, no subtitles, no other languages. how to fix that?

Does not work don't waste your time with garbage.

Fills my screen with millions of lines of garbage text. No closer to having anything usable.

Still trying to find alternate ways to convert this. Between censorship and youtube doing this, can't see any reason at all to use google products ever again.

This worked for me - when ripping from other site - however I had the following error:
Protocol 'http' not on whitelist 'file,crypto,data'!

By adding "-protocol_whitelist file,http,https,tcp,tls" before "-i" fixed this error. So:
ffmpeg -protocol_whitelist file,http,https,tcp,tls,crypto -i [input] [flags] [output]

This command is all you need: ffmpeg -i "http://host/folder/file.m3u8" -c copy file.mp4

Thanks! How to speed up the download (like, download concurrently?)

This command is all you need: ffmpeg -i "http://host/folder/file.m3u8" -c copy file.mp4

Thanks! How to speed up the download (like, download concurrently?)

@ilyazub
For concurrently downloading you can't just use ffmpeg by itself. However, here are 2 things you can try:

Option 1

Use yt-dlp** or youtube-dl with aria2c as an external downloader. This just makes yt-dlp leverage aria2c for media downloading so you get concurrency.

$ yt-dlp --external-downloader aria2c --external-downloader-args '-c -j 8 -x 8 -s 8 -k 2M' "https://host/folder/file.m3u8"

Note the args I've set for aria2c in the above example. These tell aria2c to continue partial downloads, to use 8 parallel download requests, and to limit each request to 2MB each (-k 2M). Feel free to try tuning these parameters. For most situations limiting each download to 2M will slow things down but sometimes media hosts (dailymotion) apply traffic policies that allow the first part of a download stream to go very fast, but then they start rate limiting the download as it continues. By restarting each request after 2M you can sometimes subvert these limits.

Option 2

Use yt-dlp or youtube-dl to extract the relevant URLs and then use aria2c manually for concurrent downloading. Like this:

$  yt-dlp -g "https://host/folder/file.m3u8"  # the -g flag tells youtube-dl to just return the media url[s]
...
https://media.clip.host/..../media.mp4
$ aria2c "https://media.clip.host/..../media.mp4"

Warning: often a root .m3u8 references other .m3u8 files which then reference the media so you might need to follow the tree of references a few levels deep.

**These days I recommend yt-dlp over youtube-dl. It's just a fork but it seems more regularly updated and in general succeeds more often for me.

I'm having a problem trying to decode when the player is using the fake img link, I see the video file has a ?PNG at the beginning of the content, anyone have a way to fix this?

File .m3u8

#EXT-X-VERSION:3
#EXT-X-TARGETDURATION:7
#EXT-X-PLAYLIST-TYPE:VOD
#EXTINF:6
https://s5.bestchill.workers.dev/sg/MjAyMjA5MjJfMjgg/MDAxNjYzODM5MTIyMjQz.2NlYD5uKQINWCu4tSseNg5wAEMPSpLqy1uPC7PsqwrAg.rPpnj1itZm2hbQWcxJ3uGzDHhzhWQDeirLrwZ4zrcHkg/index.html
#EXTINF:6
https://s1.bestchill.workers.dev/sg/MjAyMjA5MjJfMTM0/MDAxNjYzODM5MTIyMjA4.09L5NGYqCUWXn0Ff__jugCt3hzYxnovpjrvg0fG4SVIg.CXsKbhRODTttikxbAd0RQiQnvvWDAKtU_cNhlsieookg/index.html
#EXTINF:5.6
https://s3.bestchill.workers.dev/sg/MjAyMjA5MjJfNDgg/MDAxNjYzODM5MTM0OTc1.o-awTLVDuR2c1fV7bdavYODnsl7M0p5PCP05Q07fjiQg.8LIfxzp--TCKU3buOCkznGCGHMvnyabAbjvyH2pN0MEg/index.html
#EXTINF:6
...

File HLS index.html

�PNG � IHDR����ؿů pHYs�������� cHRMz%������u0�`:��o�_�F�IDATx�b,�����������
...

This is 2 file example:

https://github.com/tronghieu60s/tronghieu60s.github.io/blob/master/index.m3u8
https://github.com/tronghieu60s/tronghieu60s.github.io/blob/master/index.html

@varenc Thank you the solution with yt-dlp and aria2c!

if the url in m3u8 should with some header(like cookie), how to run it?

sometime some url thow 403, if I ue curl with header, it run successful

maybe we can add a Reverse Proxy between ffmpeg and site

If you run this exact command on newer FFMPEG versions you'll probably get an error about something not being on the whitelist: protocol https not on whitelist

Modify the command to use this instead

$ ffmpeg -protocol_whitelist file,https,tcp,tls -i master.m3u8 -bsf:a aac_adtstoasc -vcodec copy -c copy -crf 50 video.mp4 

Tested with FFMPEG versions 5.0 and newer

Thanks! How to speed up the download (like, download concurrently?)

Use N_m3u8DL-RE with -mt argument.

Tried using VLC and it was being tempermental on me, this worked so much better thank you

can you please share more details how we can convert using VLC Kit?

Works great also i noticed if you copy a 1280x720 youtube video url from download helper and use this method it downloads the 1920x1080 video instead it downloads a higher qualtiy video then the original. It works perfectly thank you.

This works perfectly, thanks for sharing!

nice work

You can use this approach to emulate the same headers that your browser sends, thereby bypassing 403 Forbidden errors:

Using Firefox, open Developer Tools using Ctrl + Shift + I, or click on the 3-line hamburger menu in the top-right (1.) > More Tools (2.) > Web Developer Tools (3.).

Go to the Network tab (4.), reload the page, find the appropriate .m3u8 entry, typically named stream.m3u8 or something similar (5.), right-click, Copy Value (6.) > Copy as cURL (7.).

Then paste the resulting cURL command into here: https://windyakin.github.io/curl2ffmpeg. Download and install the free, open source FFMpeg tool if you don't already have it, paste the resulting ffmpeg command into your commandline (you may have to add a .exe suffix, eg: ffmpeg becomes ffmpeg.exe if you are using Windows) and enjoy the video!