Manipulating Decibels


I’m trying to finalize audio files to submit to ACX (audiobook publisher) and I see that the mastering guidelines say:

Your submitted files should measure between -23dB and -18dB RMS with peaks hovering around -3dB.

I’m using Audacity 2.0.0 for Windows, and for the life of me I can’t seem to figure out how to make sure my files meet the above requirements.

Help, someone?

Please reply with explicit instructions as I am rather clueless.

Thanks in advance!

There is an optional plug-in called [u]Wave Stats[/u]. (There seems to be a time-limit, and you may only be able to analyze 30 seconds at a time.)

Or, you can check the peaks by starting the Amplify. And, you can enter -3dB as the New Peak Amplitude and that will make the adjustment. You can cancel the effect if you just want to check without actually changing the levels.

Amplify and Wave Stats will give you the RMS and peak… I don’t know of a way to measure peaks “hovering around” a level, except by looking at the waveform. With the default settings, 1.0 is 0dB and -3dB is around 70% (0.7) You can change to a dB display, but it’s not easy to see where -3db is. For a voice recording, I’d probably Amplify the peak to 0dB, and check “hovering around” by looking at the waveform visually.

I think most audio editors have something like Wave Stats built in. I also use GoldWave ($60 USD), and I can get the peak & average, but I have to use “tricks” similar to using Amplify with Audacity.

[u]SOX[/u] (FREE) can give you the details/statistics if you are comfortable using a command-line program.

Now here’s the tricky part… The ratio* between the peak and average (or RMS) is constant… If you boost the volume by 6dB, the peak and average will both go up by 6dB. You may need to use compression, limiting or adjust the volume envelope in order to change the ratio and hit both targets.


  • Decibels are logarithmic so the difference (subtraction) is a ratio.

There was a thread a while ago on how to tell people to meet the audiobook specification without becoming a recording engineer. Part of the problem is getting something to tell you what the RMS value of your voice is. No regular Audacity tool will do that.

[Graphic of clock running backwards] I’m remembering this now.

I did a speaking test using a radio show with notoriously wild sound levels and applied Chris’ Compressor with the compression setting (first option) reset from 0.5 to 0.6.

Test Clip:

I’m betting that one tool gets you inside the ACX specifications without you knowing what “Attack and Release Times” are. I would apply this tool to one of your performances, listen to it to see what it did, submit it and see what they say. Take super good notes if they reject it.


Thanks so much Doug and Koz! I’m definitely not an expert so I really appreciate your help. I’ll save your advice and do my best to do it justice! :slight_smile:

I’ll save your advice and do my best to do it justice!

It would be good if you read through a page or two of text, run it through Chris — alone — and submit that as a test posting just to see what they say. Reading the whole “Hitchhiker’s Guide” only to have it fall flat three weeks later at the quality acceptance step isn’t helpful.

The whole forum is waiting for the results.

No pressure.


Effect > Amplify won’t.

Analyze > Contrast… can be used to measure (I assume) unweighted rms of a mono track.

Analyze > Sample Data Export… can write the unweighted rms of a small selection into a text file.


A portion of the ACX documentation is explicit about wanting the loudness variations of a show(s) to be small so as to avoid having to constantly adjust the volume. If you recall, that’s also a page from Chris’s Compressor book.

RMS is one semi-good way to measure that. Remembering that this also the way that the widely despised ANSI-C16.5 American VU meter works.

I believe that clip I posted simply and easily conforms to the ACX guidelines, but they haven’t been spritely in getting back to me for testing. So we need somebody with an existing account (hint, hint).


It would be better to update to current 2.0.5: .



I contacted ACX about their technical specifications for loudness, peak volume, etc. etc. and it’s possible we may need a live ACX account to do testing.

I’m not interested in signing with HarperCollins to publish a book, so the full-on account pathway isn’t useful to me, but if you already have such an account and are willing to test with us, then we can tell you point by point how to produce your work for publication.

You don’t have to be an audio recording expert. We can help you.

One of the longest message threads on the Audacity forum was a guitarist in Portugal who just wanted to record his acoustic guitar. Seems simple, right? It didn’t have anything to do with Audacity directly, but we have experienced people to jump in where needed.

Let us know.

Audacity Team
Los Angeles

Yes, definitely, I would be thrilled to have your help with meeting the ACX specs. All the moreso because I’m basically just an author/publisher trying to publish my own audiobooks and I really have no idea what to do about the intricacies involved in sound mastering and so forth!

I fiddled around a bit with my ACX account last night, uploading a couple of files and trying to get the cover art right, so I have the requisite account. I was going to just go ahead and use the compressor you recommended on all my files (the audiobook is not terribly long - about 45 minutes for the whole thing), run noise removal and submit them. I like the idea of submitting test files, but I wasn’t sure if I could do that or if I had to submit the whole thing at once.

Anyhow, I’m definitely game! How should we proceed. I saw that ACX has a phone help line I could call - maybe I should start there?


More as I get there.
The account moderation (wait for posting) will vanish after some number of posts. We need to make sure you’re not trying to sell us storm windows or manhood enhancing drugs.

You’re not are you? Storm windows I mean?


the unweighted rms of a small selection into a text file.

Good point. There’s nothing in the spec about “A Weighing” or anything else. It’s straight “Battery Equivalent.”

Noise. See Above. Xina has applied Noise Removal to the show and over and above the possible problems of doing that, how do you tell if it was needed?

Assume Noise in the background of a performance. That is, ffffffff hiss from the microphone preamplifier or wherever (Let’s leave Gran Molestia , the neighbor’s Chihuahua out of this).

Since RMS is 0.707 x a sine wave peak value, and since noise is an infinite number of sine waves, can we fuzzily assume that the ACX Noise Floor is 0.707 times the (+) peak value of noise when nobody is speaking? You can get the peak value of a small section of the show by watching the green bouncing sound meter. It may be necessary to bump the sound meter sensitivity up to get a meaningful value.

There is no single noise value because the measurement subject and the noise itself are constantly changing.

I wonder if that turns out to be a simple dB adjustment.


It is. The RMS value of a wav is 3dB lower than the peak reading.

So if your Audacity green sound meter is bouncing at around -58, then your ACX noise value is -61 which passes.

That’s still pretty steep. My live recordings (with some exceptions) hover around the -48 range. That will not pass no matter how you measure it.


Noise level 50’something dB below peak level is decent, more is better (ACX recommend a peak level of -3 dB and a noise floor of -60 dB, which is 57 dB below peak). What they definitely don’t want is: “talk, talk, talk | (absolute silence) | talk, talk, talk…” because that is completely unnatural and is more distracting for the listener than having a small amount of noise (“room tone”).

What they definitely don’t want is: “talk, talk, talk | (absolute silence) | talk, talk, talk…” because that is completely unnatural and is more distracting for the listener than having a small amount of noise (“room tone”).

I don’t remember seeing that in the ACX Requirements …

I bet if you hire out your mastering, it comes back with greatly reduced or no noise floor.

They are explicit that whatever you do, all the chapters and parts of your performance match, and hopefully match all the other submissions on ACX. And in this regard, they’re duplicating the performance of broadcast radio (at least in the US). You can go whipping up and down the dial and all the shows will be “no volume control” even. If you noticed, that’s not the case with podcasts. You can get impossible volume variations within the same podcast.

Still, that’s an awfully difficult noise floor for a performer to get to in casual circumstances.


You commented that you were designing the cover art for your publication. It sounds to me like you will get all the way up to finished product submission before ACX tells you whether or not your sound is OK and they will accept your work.

So that’s the phone call to ACX Support. Is there a way to test your sound, say in segments or chapters before you throw several weeks of work into the trash bin.


And there’s another thing you can do. Walk us through the process of signing up for your ACX account. What were the screens and forms you had to fill out? You are publishing your own stories, right? You’re not reading someone else’s book?

Is any of this finished work available in a posting other than ACX so we can listen to it? The forum can host sound files, but they can’t be big enough for serious analysis.


Well. That’s awkward. Attached is what you have to do to see your noise floor. At one time I knew how to split Record and Play.
Screen shot 2014-04-03 at 8.53.36 PM.png

Sooooo close. You can almost do this with 60dB meters and a little guesswork, but for an unfortunate graphic design. If your noise peaks about where that unlabeled line is (-57), then you pass.

Screen shot 2014-04-03 at 9.01.34 PM.png

ACX have recently revised the wording on their help pages. The current version does not mention “room tone” but I discussed the issue (which was in their previous wording) with them.
This was the reply from the ACX Production Coordinator:

  1.  We prefer that editors use room tone to edit their files, as opposed to simply replacing noises/edits with dead space. The “gating” effect you refer to is precisely the sound we’re attempting to avoid, and it is also the very same sound that we sometimes end up hearing when producers send us edits performed in this way. Just keep in mind that we will only reject a title that is edited with dead space if the noise floor is so high during the spoken word sections that it is jarring to the listener’s experience (where the environment changes character radically between dead sections and spoken sections). If the producer prefers to use this method, however, and can obtain a low enough noise floor in the process, we’ll most likely pass the title. With that being said, ACX titles should have a noise floor (room tone level) that does not exceed -50dB RMS. If you can get the noise floor lower than that, great!

So we’re in 'Secret Aural Teachings" territory and it’s perfectly possible and expected for ACX to reject weeks of finished product.

Not helpful.