I'm not aware of any plugins, but in case anyone is curious about how to replicate that setup on Linux systems with PulseAudio, you can create virtual outputs with:

    pactl load-module module-null-sink \
        sink_name=Device_Name \
        sink_properties=device.description=Device_Name

If some program filters out output monitors from its input list, you can usually use pavucontrol to force-change it. Or, you can create a linked virtual source:

    pactl load-module \
        module-virtual-source \
        source_name=VirtualMic \
        master=Device_Name.monitor \
        source_properties=device.description=VirtualMic

Then in OBS settings you can set up the virtual device as Monitoring Device.

I took the commands from one of my side projects, maybe someone will find other parts of it interesting: https://github.com/pzmarzly/mic_over_mumble

Thanks! This fixes my issue!

Use Equalizer APO with a noise-reduction VST. This is by far the best option on Windows if you are concerned about latency×compatibility×underruns as a figure of merit. It is far superior to using "virtual audio cables" and a VST host (like Lighthost or SAVI) or voicemeeter.

It's not well known, but it really works spectacularly well compared to those other options. For me it never has had any audible buffer underruns (unlike Lighthost), no noticeable latency (unlike SAVI and Voicemeeter, even with small buffer sizes), no problems regarding exclusive mode (unlike voicemeeter) and it works with every single application.

The UI is not terribly clear about this, but it can drive multiple devices independently, simply by adding several "Device" blocks to the configuration.

Wow, I just did this and it's crazy how much better it sounds. It's a little complex to set up, but this guide makes it super simple: https://antlionaudio.com/blogs/news/removing-background-nois...

This was great advice! The key thing about Equalizer APO is that it does its processing of the sound before the Windows sound API, so any program recording from my mic gets the processed version, and there's no virtual microphone needed (well, the actual microphone becomes virtualised - there's no separate device). I have that all working, and mostly followed this tutorial: https://www.youtube.com/watch?v=J3fBx2ftaBs

I adjusted the numbers for my setup but his were a good starting point.

What VSTs would you recommend for improving audio quality of video calls?

ReaFIR works quite well and has very low latency. It's a bit fiddly to auto-generate a noise profile due to the architecture of Equalizer APO (the entire audio processing runs inside the Windows audio stack, so the VST panels in the configuration editor don't have a signal). Basically you use another VST host (e.g. Lighthost or OBS), generate your noise profile there and then copy/paste the chunk data into the APO config file.

Some general EQ'ing on the mic also works wonders for how well it sounds, but that's very specific to your voice and mic.

--

Another use case of Equalizer APO where it is much better than everything else is compressing game audio. Some games simply have audio that was designed without regard for hearing safety (CS:GO is a strong contender for #1 here), and this helps immensely with it.

ReaFir is amazing.

Do you have any recommendations for VST plugins to use?

Please see my sibling comment. The video I linked uses ReaPlugs (https://www.reaper.fm/reaplugs/) and I ended up without needing any explicit noise reduction, just noise gating and some other adjustments.

On Mac I use Soundflower to redirect audio between apps.

There'a also AULab which can be used for some effects in realtime (e.g. adding reverb and redirecting to built-in audio out).

https://github.com/mattingalls/Soundflower

https://www.apple.com/nz/itunes/mastered-for-itunes/

Blackhole is a nice Soundflower alternative that might be appealing for some folks in certain use cases as well: https://github.com/ExistentialAudio/BlackHole

I have no association with them, but https://krisp.ai might help you.

I think there should be software that provides virtual audio devices, so you could configure it as the output for OBS and the input for you other application - something like this: https://www.vb-audio.com/Cable/

Yes, that's what VoiceMeeter does (same author).

I use Loopback on Mac, and its never let me down! https://rogueamoeba.com/loopback/