Noise Level Question

Hey guys! Hope all is well. I have a question regarding noise level/ACX requirements. After recording a majority of my first chapter and analyzing it with ACX Check, I am receiving a warning for my noise floor being too low. What’s frustrating is that the audio sounds great. I was curious as to whether or not this warning meant that my files would be denied by ACX? For both the Peak and RMS levels, it specifically says “pass” or “fail” but for noise floor it only gives a warning. Is this something that will hinder my file from successfully being submitted or is this something that I should not worry about? As mentioned, the audio sounds great and I would love to keep in this direction if possible. Any advice/info would be greatly appreciated.

ACX has a failure they call “Overprocessing.”

Nobody can record with super quiet noise levels, so you must have either recorded with processing applied somewhere in the recording channel, or you over-applied Effect > Noise Reduction in post production. That much reduction almost always produces some distortion such as talking into a wine glass or milk jug and that will get you rejected.

Announce this ten second voice test and post it on the forum.

Side note: Announce it like you’re trying to sell us milk, not like this is boring and you really want to be somewhere else.

Don’t Do Anything To It. Record, stop, trim to ten seconds if needed, export WAV, and post it.

Right about the beginning of the pandemic, ACX peeled off their live performance quality control in favor of non-human testing. They have an on-line testing scheme similar to ACX Check. They don’t check noise.

They also don’t check theatrical presentation, so you won’t know if your voice won’t pass until you finish the whole book and submit it.

I got supersonically lucky. My voice didn’t pass, but I did it when they still offered live human analysis of a submitted short test reading.

“Your sound file is technically perfect in every way. That’s the good news. Your voice needs a lot of work.”


After recording a majority of my first chapter

ACX has a couple of side issues before they will publish an audiobook. I need to be able to buy your book on Amazon right now. The submission process has that built in and there’s no known way to work around it.

It’s best if your book is a story with plot, characters, and setting. ACX publishes a list of books they’re not interested in. Scroll down.

No cookbooks or chants.

You don’t have to be a Hollywood/Broadway actor. I have several audiobooks from Sarah Vowell. She has an odd reading style I like.


As always, I am truly appreciative of your help! After doing some research, the problem is fixed. Currently, my numbers are showing as follows:
Peak Level…-3.27 db (Pass)
Rms Level… -19.91 db (Pass)
Noise Floor… -75.65 db (Pass)

Thanks again!

For interest, what was the fix?

I found it after watching the following video. At the 5:01 mark he explains how to fix a failing noise floor.

It might be worth posting a voice test anyway. There was a recent poster who built live performance pauses and gaps into his reading and the result was very uncomfortable to listen to in an audiobook. We can’t always predict problems.

There was also a reader who had difficulty reading work they didn’t write, and one who was medically asthmatic. Theatrical reading is harder than it appears.

I’ve been recommending you listen to a published audiobook as an example instead of trying to interpret ACX’s instructions cold. You can get one from the public library.

It’s not unusual to catch problems at an advanced chapter and have to start over. Audacity is pretty good at post production processing, but we can’t fix all problems with careful editing or an effect or filter.


I found it after watching the following video.

There’s somebody in love with his own voice…

His example created some damaged work and then went about “fixing it” by boosting subsonic and supersonic noise. He talked about the older technique of injecting room tone between words and phrases. He was right. That takes forever and is prone to errors.

Neither of those techniques addresses basic recording problems and it’s possible in some cases of noise clipping for it not to work. It is strongly recommended that you find out why your recordings sound like that.