I’ve used mic placement, mouth breathing, etc, to good benefit, but I’d still like to reduce my breathing sounds in a way that ACX won’t reject.
Spectral edit parametric EQ is very effective for this. Can it be combined with a gate function, so that I don’t need to select every breath section between sentences?
Related question – does Noise Reduction respect a Spectral Selection, or does it use all frequencies within the selection?
Thank you in advance.
Zash
It seems like Combined low-shelving and high-shelving filters can act the same as Spectral Edit Parametric EQ but with a wider frequency range of maximum reduction. Though I don’t see how to gate the shelf effect, either.
I thought I was on to something with the ReaGate plugin, when I saw that the settings allowed me to combine lowpass and highpass filters and also invert gate (duck)–but after watching a ReaGate tutorial, I realized that the filters being under “Detector input” means they only affect where ReaGate applies the gate–it’s not applying a spectral edit at each gate location. 
Well, I think I’ve found a solution: using the “Gate frequencies above (kHz)” setting in Audacity’s Noise Gate. This provides the gated low-pass filtering I was after, but not the high-pass – which I don’t think I need, after all.
btw, I’m doing this because I don’t like just gating the breath sections and driving the room noise way down at the same time. That just seems like something an automated ACX algorithm would flag.
Here is a typical spectrogram of my voice.
Three sections of breathing in are visible; I’ve done a spectral selection of the left section. My breathing noise is between F1=860Hz and F2=4700Hz.
So the first thing I tried was painful… using a sharp roll-off, I filtered to three tracks: top one above F2, mid one from F1 to F2, bot one below F1. I then applied Noise Gate to the mid track (“Gate frequencies above” was left at 0). Then I mixed the three tracks and I was kind of surprised that it sounded very good, with the breath sounds tamped down, the noise floor left untouched, and no noticeable artifacts. But to do this to every chapter file, ugh.
Then it finally occurred to me that, as in the image above, there is very little energy above F2 in the breath sections to be gated (because I’m not speaking then, and the room noise is low frequency). So Audacity’s Noise Gate with “Gate frequencies above (kHz)” (basically a gated low-pass filter) should work fine–and it does.
After cleaning up the breaths and other issues (I haven’t played with Click Removal yet), I preselect my Noise Profile, and then apply my macro, which has a few more steps than Koz’ audiobook mastering macro (which may or may not be a good thing…):
I’ve been toying with different levels of Noise reduction (dB).
So thank you to the developers for including the extremely helpful Noise Gate “Gate frequencies above (kHz)” feature!
Zash
I’ve been able to achieve spectral gating with a Filter Curve EQ as discussed here.
