Need help with audio steaming from one qube to another via network

User Support
1

Need help with audio steams from one qube to another via network

I want to implement a workaround for the issue #10320 of sound of Windows 10 qube being crackling, probably due to 44.1/48kHz broken resampling (it’s a regression of R4.2 compared to R4.1).

To work-around this problem I want to stream all sound from one Windows 10 qube to another Fedora-based and play all sound there as usual.

I want a solution that would provide no sound delays and buffering, just stream sound in any codec and any compression or without compression, and play in Fedora-based qube. Seem to be completely doable.

I almost solved this, but still need help from somebody experienced, especially with “no delays” part.

What I did:

  1. Created 3 qubes: windows qube, firewall-router qube and fedora based player qube for playing sound, let’s call them: win → router ← fedora

Both win and fedora qubes have network qube (netvm) set to router qube.

  1. I adjusted firewall rules of qube according to docs:
    Enabling networking between two qubes

So, I now have direct IP connection from win to fedora and vice versa. So far so good. Pings go without problems in less than 1 ms, so sound delay should not happen due to network.

Let’s call IPs of these qubes: winIP, routerIP, fedoraIP. Essentially these are 10.137.0.* IPs, of course.

  1. I installed Virtual sound card in win qube

I used VB-CABLE Virtual Audio Device, which is donationware/freeware driver from here: VB-Audio Virtual Apps . It has good reputation, so, I am OK with running such proprietary tool (not FLOSS) in Windows qube, as the whole OS and other apps are proprietary there, too.

I set the system sound output card to CABLE Output (VB-Audio Virtual Cable) in win qube.

  1. I installed ffmpeg in win qube.

And I can run there:

C:\ffmpeg\bin> ./ffmpeg -f dshow -i audio="CABLE Output (VB-Audio Virtual Cable)" -ac 2 -ar 48000 -c:a pcm_s16le -f s16le -flush_packets 1 "udp://fedoraIP:12345?pkt_size=800&buffer_size=1316"

This command streams all sound from audio card to network address fedoraIP:12345 via UDP (12345 is just port number that I selected randomly).

Alternative way is to run OBS application in win qube and stream audio from it with GUI, but I did not try it yet. Will it be better?

  1. I have ffmpeg in fedora qube, too.

I run ffmpeg there to play the receiving stream (via ffplay in my case):

ffmpeg -f s16le -ar 44100 -ac 2 -i udp://0.0.0.0:12345 -fflags nobuffer -flags low_delay -probesize 32 -analyzeduration 0 -max_delay 0 -reorder_queue_size 0 -f wav - | ffplay -nodisp -autoexit -fflags nobuffer -flags low_delay -probesize 32 -analyzeduration 0 -

And it WORKS! Kind of.

The audio is played without clacking nor noises, finally. But the problem is the starting delay, which is like unacceptable, like 5 seconds. So, when I open video player in win qube and start playing, in 3-5 seconds I get sound that comes to dom0 (default audio qube) from fedora qube.

The sound is fine, but when I move the playing position in player, I still have to way 3-5 seconds lag for sound to be changed. Completely non-functional for my case. If one plays something like a game, the sound being delayed by 5 seconds is not acceptable.

Any ideas or experience how to make the playback change fast and instant?

I thing something like buffering should be reduced, but I am not sure how to do that.
Or maybe the problem is the chosen virtual sound card? Or maybe stream codec or something else?

1 Like
2

BONUS:

Here is an example of commands for GNU/Linux → GNU/Linux sound streaming that works well, with no delay. Let’s say we want to play some desktop-login.ogg file from sender qube to another receiver qube (with receiverIP):

SENDER:

ffmpeg -re -i 'desktop-login.ogg' -ac 2 -ar 48000 -c:a pcm_s16le -f s16le -flush_packets 1 "udp://receiverIP:12345?pkt_size=800&buffer_size=1316"

RECEIVER:

ffmpeg -f s16le -ar 48000 -ac 2 -i udp://0.0.0.0:12345 -fflags nobuffer -flags low_delay -probesize 32 -analyzeduration 0 -max_delay 0 -reorder_queue_size 0 -f wav - | ffplay -nodisp -autoexit -fflags nobuffer -flags low_delay -probesize 32 -analyzeduration 0 -

It works great, almost real-time.

I have no issues with sound from Linux-based qubes, but if somebody has it or wants some audio configuration that is more complicated than one Qubes OS has, the way to go for you can be installing additional usb soundcard, pass it to some qube and put sound streams from some qube(s) to it like that.

1 Like