I run a hybrid local and remote meeting twice a month.
Jitsi instead of Zoom, but the same idea. It took me several months to finally iron all the kinks out, and that was one of them.
I started on Windows, and then it became too much of a mess with all of the interconnections between apps that each only had a partial solution, so I rebuilt it on Linux. I've attached the documentation that I have for the audio signal flow and the startup/shutdown script. Some of what the script does is to manage the "very capable mess" that Linux's audio is at the moment, so you'll have to re-design that part to do the same thing on Mac.
---
One of the massive keys to making it work well, is to do all of the audio work in a DAW, not in OBS. In my case, that DAW is
Ardour, but you can use whichever one you like. They all work about the same.
OBS, then, sends its video soundtracks out as quickly as possible - no processing, no streaming - just get them out of there, and into the DAW. For things that you can move entirely out of OBS, do that. OBS only does video here, not audio. Use OBS's Monitor to feed a loopback that the DAW picks up. And the DAW gets things like mics directly; OBS never sees them at all.
The DAW, now, feeds everything that needs audio:
- Speakers
- Headphones
- Meeting
- OBS for streaming or recording
- Etc.
So OBS's Monitor is free to use for the purpose above. The return from the DAW to OBS, is *only* for streaming or recording. Don't Monitor it in OBS.
Again, OBS does no audio processing at all. That's all done in the DAW. OBS is just a dumb passthrough. Get everything into the DAW as raw as you can, finalize each version there, as each destination needs, and send it to all of those various destinations with no further processing beyond the DAW.
