downside of "silence audio?" Negative amplify?

I’m just getting started on narrating an audiobook. I’ve found that “silence audio” works very well for getting rid of breaths and when I listen afterwards it sounds good–not like a bad gate. It seems pretty transparent and it’s easy to use: select an annoying place and hit control+L. However I used the "analyse/contrast tool on a passage I silenced and it came out -172 dB. That seems bizarre and the drop in volume should be obvious, but it’s not, to my old ears anyway. Is this a usable method for getting rid of extraneous sounds? Or will ACX reject it?
Second question. When I “fit project” and then “view/fit vertically,” I can see a couple of peaks in my four minute sample. I selected just the peak and very little space on each side of the peak and used effect/amplify to reduce my selection by -3 dB. It sounds better to me. Is this good practice? Should I then normalize again? Would compression be a better choice here?

“Silence audio” is brutal. Not nice.

You could use “amplify” (with a negative number) on breath sounds and noise, or “noise removal”. That would take some experimenting to get right, but would sound more natural. I sometimes use envelopes to reduce noise, as it is even more gentle, with a fade in before words and a fade out after.

http://manual.audacityteam.org/index.php?title=Envelope_Tool

If you need to edit volume anyway, there’s not much extra work in it.

I don’t know nothing about ACX, but be aware that the first hurdle is just an automated test, looking at numbers. After that test, a human ear will judge it and he or she will probably fail you for unnatural sound, like absolute, brutal silence.

-172 dB isn’t even remotely possible in the real world. The theoretical maximum dynamic range of digital audio at 16 bit, 44.1 KHz, is around -120 dB. 24 bits will give you around -140 db. Very good gear will take you to -100 dB. It is an example why silencing audio completely isn’t the right procedure. It will even confuse analysis tools.

There are a couple of Nyquist filters to help with editing mouth clicks, breath sounds etc. But I’ll let other forum visitors comment on those as I haven’t played around with them enough to be a real help to you.

To quickly separate failures from the tons of newbies reading audiobooks, ACX has a “robot” that blindly checks for basic sound values.

https://forum.audacityteam.org/t/measure-between-23db-and-18db-rms/32770/16

There is an analysis tool flynwill developed that you can install and it will tell you all the numbers on your show in one pass.

http://wiki.audacityteam.org/wiki/Nyquist_Analyze_Plug-ins#ACX_Check

To your specific point: you can’t just beat your performance bloody and force it to be “robot compliant,” because the very next thing that happens is Human Quality Control. They have a failure called “Overprocessing.” That will crash the submissions with honky, wine-glass voices; compression artifacts; odd, pumping background noises and excessive voice clicks and pops. If you’ve been Control-Lling your voice clicks, you may have odd sounding holes in your Room Tone or background sound.

Are you recording in 44100, 16-bit, Stereo; or maybe Mono, or 32-bit, Mono? Your Analyze > Contrast numbers may come out funny if you’re trying to record in much too high sound standards.

Almost all new posters have trouble with their “studio,” giving them a seriously high background noise for any of a number of reasons.

If you feel like it, record a forum test clip and post it. Only post the raw sound file. No processing.

http://www.kozco.com/tech/audacity/TestClip/Record_A_Clip.html

Koz

I’m recording in .wav, 16 bit, 44.1. When I use analyze/contrast on my raw files I get around -35 dB for speech and -65 for silence, for a difference of 30. When I’m recording the bar graph shows -50 silence and -12 peaks. Is this good enough? I’m using an AT2020 condenser through a Focusrite Scarlett interface and I’ve treated my room with acoustic foam. I have to edit out the occasional dog bark and car muffler. I believe that going forward I just need to get rid of the breath sounds. Then if possible figure out a mastering plan.

Those numbers seem OK.

If in doubt, post a short, untreated part to the forum.