Improving microphone quality in Linux

Hey all!

So I’ve noticed a common complaint among those I interact with on Discord, lately. Whenever I’m booted into Linux they say I sound awful, the fan on my desk is too loud, and I sound like I’m calling from an old cellphone. I’m using a USB condenser
microphone made by TONOR.

When I switch into Windows 10 (still dual booting only for COD:MW) they say I sound much, much better.

I noticed that Discord offers ‘Krisp’ for the client in Windows, but not for the Linux client. I think this is the source of some of my particular woes, but I got curious to see whether or not there were any ways I could improve the quality of my microphone in Linux.

In the past, I have meddled with Jack with some success, and I was considering going back down this route (supplanting Pulse Audio). My question is thus: does anyone know of a good starting point to improve microphone quality on Linux?

I’m curious if Linux has some basic secret sauce for improving audio input but if you’re hardware requires software editing to make it acceptable you may be better served fixing the hardware first.

I’m just using an Edutige ETM-008 shotgun mic on a 3.5mm extension to my motherboard audio but on systems with a lot of electrical noise I use a mini Soundblaster USB to 3.5mm with it. The feedback i’ve gotten is pretty good. The whole package is so small you could even put it in your wallet.

1 Like

You’re the second person to recommend pipewire. I am intrigued, but instead I went another route. I edited my ~/.config/pulse/daemon.conf file to add some specific settings like resample-method = speex-float-# and avoid-resampling = true to try to improve it. I made several other changes and can post it here, and when I compared the results of my mic (by ear) recorded in OBS Studio, I thought my config changes made a difference.

I’m happy to copy over my config if anyone wants it, but it’s a pretty big config file (plus my default.pa file in the same folder).

1 Like

You could also give NoiseTorch a try: GitHub - lawl/NoiseTorch: Project dead (security breach)

Jason Evangelho from Linux For Everyone found it, but I haven’t personally used it yet :sunglasses::+1:

3 Likes

This is a fantastic thing I never knew I needed! It makes total sense, though, in light of my gripe with the Linux version of Discord not using sound suppression by default.

I’ll load this up tonight and give it a try!

1 Like

Just have to say that the NoiseTorch is my new best friend. Thank you!

1 Like

The use of CARLA may be what you are looking for in regards to real-time audio manipulation.

3 Likes

I almost want to use carla as mind mapping software.

2 Likes