What's the correct term (and solution) for this recording issue?

Windows 10. Just updated to Audacity 3.0.4 today. I have previously been able to successfully record very basic book narration using Audacity without issue. Recently, I launched into another project and I’m having issues I can describe but don’t have the correct term for. The issue happens When I first start speaking, or when I’ve paused for, say 2-3 seconds or more. On playback through headphones, Audacity is apparently failing to record the initial sound of the word I’m speaking. Obviously, I’d love to know what the issue might be and how to resolve it. (I did try 1. updating Audacity and 2. to search the manual and FAQs, including “How can I record without small skips (dropouts) or duplications?”, but each was unsuccessful and/or not seem relevant to me.) Thanks.

A noise-gate can chop-off the start of words.
(There could be a real-time noise-gate acting before the sound gets to Audacity).

Thanks. To my knowledge, I don’t have a real-time noise-gate in place. What would be an example of such a noise gate?

Also…if it is a noise gate issue, I’m uncertain why it would eliminate sounds only under certain conditions (specifically, after relatively longer periods of silence) and not others. If I just keep talking…or if I ‘prime’ the situation by saying the initial word twice…I can eliminate the problem altogether. But that leads to the need for additional editing…and tends to mess with the flow of my narration.

Hi lllighty

Have exactly the same problem with beginning of a recording (spoken word) and a millisecond chopped off.
After pressing RECORD Try doing a count-in (1234) then start your narrative.

I have a lot upper respiratory problems that cause laboured breathing that will be pronounced in a spoken word recording.

To reduce these I use the built-in noise gate. I use the ACX benchmark for my recordings. Have included in this post several Jpegs of the procedure I use.

FIG_2 - Straight after recording. No effects or processing applied yet. The first step is to apply the RMS Normalizer set to -22db
FIG_3 - RMS Normalizer now applied, but has also raised the level of my breathing-in as well (highlighted). I used the Noise Gate’s ANALYSE NOISE LEVEL to measure the PEAK value.
FIG_4 - The NOISE GATE has suggested a setting of -27db
FIG_5 - Shows I have adjusted the GATE THRESHOLD (DB) by the recommended -27db.
8db is how much I want the the breaths reduce by.
ATTACK / DECAY of 10ms is a very fast reaction time for NOISE GATE to kick-in and kick-out.

These settings work for me. Try these settings as a starting point for yourself. Yours will be different, but would only need tweaking by PEAK and how much LEVEL REDUCTION you want reduce them by.

To apply the NOISE GATE - highlight the track and click ok.
FIG_5 -
My ACX readings before
Peak level -9.58 db Warning too low
RMS level -29.16 db fail too quiet
Noise floor -61.04 db Pass

My ACX readings after processing:
Peak level -3.53 db Pass
RMS level -22.22 db pass
Noise floor -61.82 db Pass

The issue happens When I first start speaking, or when I’ve paused for, say 2-3 seconds or more.

It does sound like a Noise Gate.


Do you have Audacity > Transport > Transport Options > Sound Activated Recording turned on? That pops up every so often and performers will be missing words or portions of words.


Hi kozikowski

In my own case I think its a soundcard thing (its just cheap). The count-in seems to work every time. Because of my breathing problems, every take I do is different.

The built in noise gate, is the best thing since sliced bread.

My work-flow is:

  1. RMS Normalise
  2. Noise Gate
  3. Limiter / soft set at -3db

ACX checking throughout.

ps. Thanks for posting about ACX in another post, which I enjoyed reading and prompted me to adopt the standard.

My work-flow is:

  1. RMS Normalise
  2. Noise Gate
  3. Limiter / soft set at -3db

Gate to what? There is a caution about using Gates for ACX Submission. They like natural background noise or room tone in the background, not Blackness of Space silence. Also, gates can give pumping room noise and cause inter-word silences to not match inter-sentence and inter-paragraph silences.

ACX has a failure called ‘Overprocessing.’

Did you notice you fail ACX Check if you using aggressive gating? “Background noise is too low.”

Screen Shot 2021-09-24 at 3.34.25 AM.png

prompted me to adopt the standard.

If you mean ACX Mastering, there is another caution. Take the steps in order, don’t add any, and don’t leave any out.

Almost all home microphones produce rumble, thumping, and subsonic trash (because it’s expensive to get rid of it).

Some microphones produce such loud rumble that it competes with the voice (and it’s evil because you can’t hear it). It throws off Loudness Normalization and can give unstable production results. You might find if you use the published mastering order that rumble vanishes, your presentations get louder and clearer, and some of your P-Popping may go away.

You can gate later if you want, but probably not before. One of the features of ACX Mastering is the ability to start with almost any raw recording level and process to the right loudness. Gates can’t handle varying levels.


There is a process caution.

Don’t change anything in the middle of a book. If you discover a New and Improved method of voice processing in the middle of a reading, you can use that on the next book—or start this one over.

ACX wants everything to match.

Obsessive Engineer wants you to stop doing updates during a book. Count the forum posts where a production stopped dead “for some reason” right after a Windows Update.


Hi Koz.

Sorry, didn’t make myself clear. My croaky voice is defiantly not one for producing audio books.

I think ACX is a great standard to aspire to, even if you don’t actually use it for audio books.

My settings on the noise gate I found by trial and error, worked around the problem by looking at what I want to remove in the waveform - namely my breathing intake before a word.

I took me a couple of weeks - try this setting - trying that.

Works for me.

Noise gate can be part of “Audio enhancements” applied by Windows, (should turn those off in playback [u]&[/u] recording tabs)
or similar audio enhancements applied by software like RealTek/DellAudio. ( Such “Enhancements” can be on by default).

Noise-gates have hold times: they don’t close until the signal drops below threshold for a (used defined) period of time. If there is no “silence” for a (used defined) period of time the gate will never close: always held open.

Thanks everybody! Much appreciated! I will be working the issues based on your input over the next 24 hours or so. We’ll see what happens.