[pulseaudio-discuss] How to combine microphone audio and music from application to pipe to video conference

[Sean Greenslade]

On Mon, May 18, 2020 at 03:59:45PM -0400, Samir Parikh wrote:
> Sean Greenslade wrote on 5/17/20 10:04 PM:
> > On Sun, May 17, 2020 at 07:01:25AM -0400, Joe wrote:
> > > Someone should write a small book on this. I bet it would sell.
> > 
> > Hah, you know, typing up this response has made me consider doing a
> > proper write-up for my much more elaborate/crazy/overcomplicated setup
> > involving network audio, pulse, JACK, carla, and calf LV2 plugins to do
> > automatic livestream background music replacement with gate-triggered
> > music ducking.
> > 
> > https://dumbpic.link/manu/2020-05-17_unmufflegate_carla_flow.png
> > 
> > --Sean
> 
> If it can clearly explain the basics of Linux audio networking, the concepts
> of Pulse, JACK and how they differ, and how to do things like in my original
> question, I'll be the first in line to put down 10-20 USD for a copy!

So, I did end up doing a writeup of my more intense
Pulse/JACK/Carla/Calf setup. It's definitely on the complicated side,
and not at all necessary for solving the original quesion posed in this
thread, but if people are interested:

http://seangreenslade.com/projects/2020-05_stream_audio_processing/

On the topic of Pulse vs. JACK, I simply don't have much experience
using JACK. In my limited experience, JACK is missing some of the
desktop-focused convenience features that Pulse has, but JACK is more
focused on realtime performance and low-latency, pro-audio workflows.

In the post, I didn't go into much detail on the network audio side of
things, mostly because the solution I settled on (Pulse native TCP)
works remarkably well with basically no effort or tweaks on my part. If
you're in a totally Linux-based environment, Pulse TCP is the way to go.
I had a few false starts with RTP/RTSP and multicast, which turned out
to be very fiddly to get working reliably.

--Sean