I want to try to record audio using my mic (a Blue Yeti Nano) but I keep running into problems with noise. The mic seems to be able to pick up everything in a small radius, even my keyboard, mouse (I’ve positioned both the keyboard and mouse behind the mic, but it can still pick up the noise), computer fans, and air purifier. This is after I used the button on the mic’s back to set it only to pick up audio in front of it.
I’ve tried using NoiseTorch, but it seems to chop off the start and end of some sentences. Easy Effects seems to only work in one ear right now, and I don’t know of other software that can help make cleaner recordings.
I do know I can use Audacity’s Noise Reduction effect to reduce background noise, but I don’t want that to be my only option.
The only other option (to my knowledge) is an advanced setup where you use the linux-rt kernel coupled with apps like patchage or helvum to route your audio connections in the exact way you want them.
I use noisetorch myself, and on occasion, it doesn’t work properly even if I adjust the one setting that matters. Usually, after rebooting, it’s back to being perfect. This is, of course, annoying to say the least. Inefficient at best.
Note: Whenever I am already in the middle of work and can’t reboot, it usually helps to get a little closer to the mic, while also lowering the threshold.
It is best practice to avoid recording noise instead of recording it and try to filter it out by postprocessing. I use a back attenutor plate behind the mic and also on the ceiling. Speaker should move as close as possible to the mic. Fan from noise and other sources must be eliminated at the source. The best choice for a mic is a headset with mic.
I don’t know this specific mic, but i had some weird experience with audio interfaces that have specification for Windows drivers. Linux provides only standards support, but they might require very specific, nonstandard functions built in the driver. This might be switchable attenutors or switches for other settings, or built in filters. You are better off if you pic a microphone that at least has linux/alsa in it’s compatibility list.
Maybe the next time I buy a mic I can consider this, but for now, NoiseTorch does a perfect job 80% of the time.
It used to be 100%, but since these KDE and Gnome updates, everything is being affected in both bad and good ways. Can’t wait for this stage of rapid development to be over, especially since I use neither KDE nor Gnome on Arch.
It’s actually annoying that because those two are getting updated, they affect pretty much everything Linux-related since the libraries need to be updated.
Best solution I found is setting up a filter chain in Carla [1] with a gate to cut out the noise floor (in my case a lsp gate with a hysteresis [2]) and deepfilter [3] for all the rest (fans, mouse, keyboard, birds, …).
The disadvantages: It isn’t a one click solution and it probably takes an hour or two of figuring everything out. Also deepfilter is going to consume some CPU cycles.
On the upside: in my experience the quality is much better than NoiseTorch or EasyEffects. You can now use it in every application, not only post production but also e.g. live chatting. It’s easy to add other audio plugins into the chain e.g. if you want gain, a compressor, …
Simple setup with a virtual mic that I connect as source to the actual applications:
Here’s to hoping no Linux newbie or Windows/Mac user sees this and thinks, “Wait. I have to do all that just for noise cancellation on Linux?”.
But, in case they do see this, NoiseTorch/EasyEffects is usually enough. And if you have a mic with built-in noise reduction, you may not even need it, just like @kmonster said.
The issue right now is that Linux, in general, is under heavy and rapid development, so Arch distros will experience all the bugs that come with this. I have Ubuntu Studio installed and don’t have the same bugs I do on Arch, since Ubuntu, Debian, and other LTS distros don’t receive rapid updates.
The issue is that the Blue Yeti is a condenser mic, they are supposed to pick up that kind of sound. Otherwise you should probably go for a dynamic mic. So it’s not a Linux issue in particular, but the hoops you have to jump through to use that kind of mic in a noisy environment.
Found out pavucontrol has a latency setting today. And it seems to actually work. I currently have most of the officially supported kernels installed, but none of them auto-fixed what the most recent updates caused. But pavucontrol did?
Here’s a screenshot of what I mean. Noise cancellation is back to perfect again.
PS: Updates over the weekend messed up noise cancellation, so I decided to install three kernels: linux-zen, linux-rt, and linux-rt-lts. None actually made a difference for NoiseTorch even when I eliminated most noise and adjusted the activation threshold.
Update: It seems that because this setting is applied system-wide as opposed to being applied to one application, it uses lots of resources. So, this is not a fix for more than maybe 4 hours of use. My device went back to its speedy self once I set it back to 0.00.
