Fixing an ACX Issue

On a recent novel I completed, ACX bounced two chapters with the following notation:

• Issue: The production includes files that contain noise at the start and/or tail end. The first 0 to 2 seconds, and the last 1 to 5 seconds of each file must revised to ensure none of them contain any noise, loud breaths and/or mouse clicks at the start before narration begins and at the end after the last spoken word.

In reviewing the sound track for the particular chapter, there are no VISIBLE indicators as to noise, no bumps, no squiggels, etc.

I’ve attempted today to fix it. Short of re-recording, what I’ve done today is deleted all the dead space at front and end and , using cut and paste, replaced it with (hopefully) silence. I just sent it back to ACX.

Will that work or is there a better way???

Some audible sounds are not visible on the waveform unless you zoom-in.

It’s possible to have bassy thuds in the recording, but not hear them because ones speakers/headphones are not capable of reproducing them, (They would still show up on spectrogram view).

Will that work or is there a better way???

Production Silence isn’t dead silent. They have a phrase “room tone,” the sound the room makes when the performer isn’t making any sound. That’s what is supposed to be in there.

If all the other chapters passed, you can borrow the room tone from one of them and paste it in place of the bad segments.

You can step your way down the rings of Hell, here. You can’t edit the MP3 you submitted to ACX. You should edit the WAV file master of the book chapter. It’s not obvious you need to make WAV files of all your work, but you should.

Remember when you made the MP3s, they had to be constant-bitrate, 192 quality? If you edit an MP3 chapter and resubmit, it won’t be 192 any more. MP3 is convenient for listening to music at the beach or submitting to ACX, but not for audio production or editing.

We’ll see what they say.


There are other ways to measure this.

Drag-select some of the offending work (even if you can’t see any blue waves). Analyze > Contrast > Measure Selection. That measurement should be -60dB or quieter (ACX). We use -65dB (even quieter) as a recommendation rather than what happened to you, where you’re so close that some chapters didn’t make it.

You can drag-select your offending area and just run Analyze > ACX-Check again. Ignore the first two values.

You can noise reduce use what you have, but that can be a little more dangerous. Drag some of the bad sound > Effect > Noise Reduction > Profile. Then carefully drag-select the part you want to reduce > Effect > Noise Reduction: 6, 6, 6 > OK.

Measure it again to make sure. Make sure you still pass the two seconds at the beginning and five seconds at the end of Room Tone.


Koz, the original chapter (v2) in question overall has a noise floor of -79.9 so they must be referring to the one second clip at the beginning and the 4.5 piece at the end.

So I ran an ACX check on just the 4.5 section at the end of v2 and it showed a noise floor of -68.6 … Peak was ok but now said the RMS was less than ACX specs (which I assume is correct since there’s no audio on that short clip). I attempted to do the same analysis of the 1 sec clip at the beginning, but that was insufficient for ACX check to work

So, today I did v3 by adding silence to front and back and deleting the old. Now v3 shows an overall noisefloor of -69.1. Then I measured the tail end (the 4.5 sec clip) and it came in at -67.4 … basically the same as in v2

Soooo … like you say, let’s wait and see, I guess. If you want to take a listen, lemme know and I’ll send you the MP3 file. Maybe like you said, there’s some invisible sound on that track

Thanx for your help, Tom

You do have to be careful what the tools are telling you.

If you’ve been doing this for a while, you may have the old ACX Check. It had restrictions for show length and is no longer recommended. The new ACX Check is here.

As you noticed, if you turn ACX Check loose on Room Tone, the first two values, Peak and RMS, will be meaningless and only the Noise value is good. If you measure a whole chapter, ACX Check will go through and find the quietest half-second. So you can’t make assumptions about the noise of specific areas that way.

If your show does have well-behaved background noise, the tool can measure it anywhere and the reading will be valid. If you start fancy editing, cutting, and messing with compression or other dynamic volume managers, all bets are off. Once you start that, it’s possible to get too quiet.

The goal is to sound natural. “Natural” is not speaking over the Blackness of Space.

You did say a couple of odd things about your chapter standards values. Are you using the published ACX Mastering Suite?

This is the short version.

You don’t have to. This isn’t the only way to meet ACX, but it does work pretty well and several posters are using it professionally.


Koz, I do use the mastering scenario you outlined above.

I’ve completed four books for ACX with none of them being bounced for any reason … but book 5 has this one chapter that I can’t get through. If they reject the fix I sent yesterday, my only recourse is to re-record it.

Koz, one more question. When I do the mastering, and I use the Noise Reduction effect, what I’ve sometimes done is repeat the process thereby often getting the noise to around -80. For example, the original read might be -60 … I then use the Noise Reduction and it bumps it up to, let’s say, -72. Then I’ll apply the Reduction again, and now it’s at an even higher number, for instance -80

Is this a good idea ???

Is this a good idea ???

No. Excessive noise reduction can harm voice quality. The process isn’t “free.” Your voice starts getting “tight” and wine-glassy.

The ACX standard is quieter than -60dB. We recommend quieter than -65dB and anything in the -70s is good to go.

Are you using a noise reduction of 12? That can cause voice quality issues. Certainly no reduction values higher than that. If you need extraordinary noise reduction you should to go back and fix your microphone or recording system—or studio.

If you’re careful about live recording you should be able to hit mastered noise values in the -63dB to -65dB range. Add Noise Reduction of the Beast, 6, 6, 6 and that gives you -70dB. Done. Go make coffee.

Maybe as high as 9, 6, 6.

You can do what you want, but certainly that second application is causing voice quality damage. Remembering that the goal is plain, natural, story-telling voice, not robotic or mechanical voice.

ACX has a failure called “Overprocessing.”


Great answer, Koz … noted and will comply

Aaaarrrgggghhh … still don’t know if they’ve accepted my fix to Chapters 3 and 7, but yesterday they bounced Chapter 1 for the same reason.

That’s a little confusing to me as wouldn’t they have detected an issue in Ch 1 BEFORE they noticed something in Chapters 3 and 7 ?

Further, I ran Chapter 1 though the Audio Lab and the Audio Analysis sections and the chapter passed with no issues … so I’m wondering if the QA is something over and above the Audio Lab analysis … and if so, isn’t the feedback from the Audio Lab and Audio Analysis tabs kind of meaningless?

Color me frustrated … I’m reluctant to waste my time trying to fix Ch 1 if I don’t know whether my fixes worked in Chapters 3 and 7 … I don’t need to repeat my mistakes

FYI, I just sent ACX this same question. The first four books i did, all went thru QA without a hiccup … now book 5 is driving me nuts/ They say expect a response in 1-2 days

I just sent ACX this same question.

There is nearly zero chance ACX will reply. They’re not in the business of troubleshooting your studio.

wouldn’t they have detected an issue in Ch 1 BEFORE they noticed something in Chapters 3 and 7 ?

I can’t prove this, but I suspect they have “Reputation Ratings.” My rating, for example, would be: “This guy has the technical performance down cold, but he can’t read. Pay attention to his tongue ticks and mouth noises.”

Yours might be: “This reader has submitted noise problems in the past. Pay attention to chapter background noise, odd damage, and make sure all the chapters match.”

Nobody said ACX was going to reject you for low noise. Super low noise signals to ACX that something was done to the performance that may cause distortion or damage. Their radar goes up. Nobody can record directly into -90dB noise without shenanigans.

Worse, now you have a reputation for trying to glue and patch a broken performance back together.

Post the ACX reply. Exactly.


Koz, I had a reply from ACX telling me to reply directly to the rejection email in order to reach the Engineer who reviewed the chapter(s) in question and so, this morning, I sent the below note:

"Color me frustrated. I did four books, all of which went thru QA with nary a hiccup. The fifth book, however, has now bounced 3 times, all for the same reason albeit in different chapters. Initially, Chapter 3 bounced, I fixed it (I think) and returned it. A couple of weeks later, Chapter 7 bounced … again I think I fixed it and returned it. Then yesterday Chapter 1 bounced for the same reasons. In all three cases, it claims noise somewhere at the beginning and/or end of the chapter.

I’ve run all the chapters through the Audio Lab and all pass with no issues found. The same with Audio Analysis tool … no issues found.

I’m confused because I would’ve thought that if there were an issue in Chapter 1, I would have been notified BEFORE Chapters 3 and 7.

Further, since I THINK I fixed Chapters 3 and 7, it would be beneficial to know if that’s correct … particularly before I use the same method to correct Chapter 1, only to find out that the methodology for making the correction didn’t work.

I could use and will appreciate any help you can give me on this. The ACX book title is TRUTH, LIES AND TROLLS "

Soooooo, now I’ll send you a copy of whatever comes of this request to ACX …

Thank you, but I’m interested in reading their last rejection.

They don’t publish their detailed analysis process and the only way we know what they’re thinking outside of their publications is detailed parsing of their rejections.

I did four books, all of which went thru QA with nary a hiccup.

It’s possible you may get special treatment because of that. Your reputation until then has been short path to publication. It was a nasty surprise when the fifth reading failed simple testing and even worse when they started looking more closely and discovered other problems.

They usually do one layer of analysis. In my case drink apple juice and practice not smacking your tongue so much. But when I tried to get further dialog the connection closed. That was before The Sickness. If you didn’t get a good detail in the rejection, then I suspect they stopped giving it unless you ask for it.

They greatly pared back their services. I think I mentioned they no longer offer short chapter human review. Your choices are submit a book or don’t.


Koz … room tone. I think you said earlier to find some room tone on a previous section that had passed ACX QC and then cut and paste that in any new recording. Did I understand that right?

Here’s what I\ve been doing previously … when I open Audacity, I’d make sure furnace, etc, is off … then I’d hold my breath and hit ‘record’ for about 5 secs. I’d do that each and every time I began a new recording. So my question is, I guess: Can I cut and paste a previously good 5-10 seconds of ‘room tone’ and use that on each and every recording by pasting it in?

Sorry I’m so dense but I confused on noisefloor and room tone … are they not the same thing?

The other question is: Once I begin the mastering, and I go through all the steps (filter curve, RMS Normalize, and Limiter) and do an ACX check and, let’s say, I get a -65db reading. Is that enuf … or should I do a Noise Reduction anyway? In other words, is the Noise Reduction a step we should always use?

No reply from ACX yet as to the base question … Merry Christmas, Koz … thanx for your help

are they not the same thing?

“Room Tone” is the noise the room is making if you’re not making any sound. That’s the traffic outside, refrigerator, etc. All the stuff before the microphone. This stuff can be a little magic. I have a bass cabinet that doesn’t go off when I turn it off and it injects a very low volume [mmmmmmmm] into everything I do. That took a while to find. I was up to Space Ship Conspiracy Theories. I have power lines over the house. I wonder if when they get wet…

Noise Floor is all the noises that the microphone and microphone system are making. Microphones make noise, some more than others. Your job is to be way louder than the microphone noise without overloading anything. You can measure that in Audacity. Audacity has a natural noise floor of -96dB. Guaranteed that almost everything you do is louder than that.

Background noise is a generic term from the point of view of the performance. What does it sound like between sentences when you stop talking and breathing.

let’s say, I get a -65db reading.

I would say that was absolutely normal for a home reader. This is where Obsessive Koz kicks in and wants to apply Noise Reduction of the beast. 6, 6, 6. We have determine that a gentle reduction like that produces a well-behaved, quiet background noise for your reading and perhaps most importantly, ACX can’t tell what you did.

That gives you a submitted performance noise of -71dB. Perfect.

I need to drop for a while.


That’s -65dB after mastering, right? You may find that the mastering tools can help your noise way more than you think. Many home microphones naturally produce a low-pitched rumble sound that can mess up noise readings. Low Rolloff sucks that rumble right out of your performance. It’s one of the reasons you must apply the three mastering tools in order.

This is where actual audio editing comes into the picture. Most of the time you should try to suppress mouth noises, tongue ticks, mistakes etc. You can cover them up with the Room Tone segment you captured along with your live reading.

The process is not Copy and Paste. You can use Punch instead.

Punch was designed to be a gift from the angels. Drag-select an error in the performance. Punch. Punch will tailor the replacement sound so it exactly fits the selection covering up the error. Under certain conditions Punch can form a gentle fade on both ends of the correction so to be as unobtrusive as possible.

I’ve never used it. Let us know how it goes.


Koz, I’m searching for the holy grail on Christmas morning. Given our issue with the first sec and trailing 5 secs which, allegedly contain noise (but unseen on the sound track), what about this for a suggested workflow:

First, independently record about 10 secs of quiet ‘room noise’ … then save this as a wav file for use later, again and again
. open a new track and do the recording
. then copy the room noise file (above) and paste in the beginning and end of the recording
. then do the mastering (filter curve, RMS, Limiter, Hum Remover and Noise Reduction
. review the entire track and remove all visible clicks, etc
. then save as a wav file Edit Master and export as an MP3 file

If this is the way to go, then I’m assuming I could always just use the recorded room noise above and paste into any new document.

Does this make any sense or am I off base?
Thanx, Tom

Also, Koz, earlier you said you wanted to see exactly what the rejection msg said, so here it is:

The ACX Quality Assurance Team has reviewed your recently-completed audiobook, "TRUTH LIES & TROLLS (Bridge Club Book 4) ". We have found 1 issue that must be corrected before we are able to process your audiobook for retail sale.
Below, we have listed the requirements not met and our recommended solutions. When making revisions, always be sure to revise the originally recorded WAV files within your preferred editing software. We recommend revising all files as this helps to instill good editing practices and ensures all errors have been addressed.
• Issue: The production includes files that contain noise at the start and/or tail end. The first 0 to 2 seconds, and the last 1 to 5 seconds of each file must revised to ensure none of them contain any noise, loud breaths and/or mouse clicks at the start before narration begins and at the end after the last spoken word.

Koz, the reason this is so irritating is that what they’re saying (contain any noise, loud breaths and/or mouse clicks at the start before narration begins and at the end after the last spoken word) would also seem to imply that these should be visible on the sound track … whereas on my sound track there are NO visible hiccups. Thus my earlier idea about recording/saving a room tone wav file and then pasting that in on every recording at the start and end prior to mastering.

seem to imply that these should be visible on the sound track

You inferred something they didn’t imply.

The normal setup of the timeline blue waves only shows you the loudest 25dB of the show. From -25dB (approximately) to 0dB. That’s where most of the production tools work and that’s where most of the production decisions need to be made. This has its limitations, as you’re finding.

You can force the blue waves to show you everything. Control+Click the vertical bar to the left of the waves > dB.

Screen Shot 2020-12-25 at 9.27.26 AM.png
You may find it handy to grab the bottom of the wave display and pull it taller so more of the numbers are visible. You may also find it handy to put it back to Linear since it’s really hard to adjust your voice volume in dB.

Look at the bouncing sound meters. What does the far left side say (you may need to guess as the actual number)  -60dB? I use different meters.

You can grab the grab bar on the left of the sound meters and put them anywhere in the menu bar you want.  Mine are on the top-left. Grab the right-hand end of the meters and pull them longer, all the way to the right side of the Audacity window.

Audacity > Edit > Preferences > Interface > Meter Range > 96dB.

![Screen Shot 2020-12-25 at 10.02.24 AM.png|1480x273](upload://qFsWakD4X9d7s5qI3GJqFFQv8lA.png)
Note with -96dB on the left and the large meters, -70dB, -60dB, -6dB, -3dB, and all the other critical sound measurements are clearly visible.

There's one other change, too. Right-Click the recording meter > Options > Gradient.
Right-Click the playback meter > Options > RMS.

That gives you recording meters which change color when they get close to the critical loud recording and overload points, and two-color playback meters which show you peak and RMS. Remember RMS?  ACX's loudness measurement?

I'll understand if your eyes glaze over and you fall down in a faint.

Keep saying to yourself, **"I am the recording engineer."**