Measure between -23dB and -18dB RMS.

Recently has updated the requirements for files submitted as audiobooks. One of them is Measure between -23dB and -18dB RMS.

I need to know how to be sure that is what I am doing?

The “contrast” option in the Analyze menu will measure the RMS of a selected section of audio …

Select the audio you want to measure and click on the button I have coloured green, the RMS value will appear in the box I’ve coloured yellow …
contrastdialog, coloured-in.png

[ you don’t need to do foreground and background measurements if you just want to measure the RMS value of a section of audio ].

Thanks for the help.

The obvious next thing is what if it is not in the range requested.
What is it that determines the RMS?

I know this is clear to many, but I’m just a guy at the mic.


The RMS value is an average value of the amplitude (level) of the audio waveform.
RMS is proportional to the total area between the waveform and the zero line , (the red area )

If you change the amplitude of the waveform using amplify* or using dynamic-range-compression it will change the RMS value.

[ * it is possible to “amplify” by a negative number and reduce the amplitude , and reduce the RMS value ]

You are very helpful.

I will simply need to check my compression and make sure I get the results needed.
It seems there should be a predictable setting. Threshold ratio etc.

Time to experiment.

Thanks again.

It seems there should be a predictable setting.

The reason the company is so particular about that value is because it’s typically not. Live vocal performances can have wild swings in peak and RMS values. If you’re used to listening to the radio or purchased performances, those have been heavily processed so the loudness values are constant.

One of them is Measure between -23dB and -18dB RMS.

Is that all it said? You can produce a vocal performance with wild, energetic swings in volume and have the average of loudnesses come out to -21dB RMS. The show would be almost unlistenable. Not that you do that, but I’ve heard performances that do and the company has to cover their bases.

Are the requirements in such a form that you can post them here? Alternately, do you have the web page address that publishes the specifications?


This is all it said. There may be more help, but I can’t find it.

I’m sure with many new indie authors doing their own narration there must be a wide range of issues on the tech side. Not to say that I have id down.

Is it true that with a particular ratio you will get rms close to what is needed? Not over compressed or under done?

It doesn’t listen to the performance and decide what to do with the show. It’s an input/output thing. If you put in this volume, you will probably get that out. If you’re too far off, you can still slide out of range.

I just know you wanted to get into the fuss and bother of this audiobook thing. One of our crosses to bear is the lack of information from the companies.

Chris’s Compressor is an outside source program that does look ahead and tries to intelligently adjust volume and range depending on content.

Chris designed it so he could listen to dramatic opera in the car. Chris got tired of cranking up the volume to hear the water maidens hiding their lump of gold and then being blown into the rear seat by the high-volume entrance of the Captain of the Guard (to borrow two different operas).

It’s been my contention that it should be a lot less fuss turning Chris loose on a show than trying to tune Effect > Compressor for different performances.

If we could get the companies to tell us what the goals are.


You should try the Compressor thing. If you have an even announcing voice, you could settle down with a appropriate suite of settings like an comfortable old shoe.


Oh, the electrical thing. You know when you see a fluorescent light flickering and carrying on because there may be something wrong with it or it’s too cold. That’s the peaks in the power waveform. Now stand in front of a space heater to warm your hands. That’s RMS. Slow heat energy.

Nobody will be warming their hands in front of your storytelling, but the instruments that measure both electrical waves are very similar.

Drum solo versus that low organ note that moves your shirt?


You will need to juggle with the compressor and the Amplify (or Normalize) effects to get both the peak level and the RMS level within the required ranges.

Thanks so much for all the above, it;s really helpful.

I have used compressor, and then amplified but then normalized so peak is -3DB but everytime I check after normalizing the RMS has been reduced. Don;t know how to keep peak and -3DB and keep RMS between -22 and -18, do I just keep fiddling around? Or is there anything else I can do?

Two years?

I have a step by step process for this, but we need to wait so I have access to a real keyboard.


Open a segment of the show before you did anything to it. The raw performance.

Do you have ACX-Check? Highly recommended. It’s an analysis tool that reads you all the ACX specifications in one shot. I have an analysis page that tells you all the things you have to do to get the same information without acx-check. It’s not pretty.

Once you have acx-check running and the raw performance on the timeline, Normalize so we can start from somewhere predictable. Modern microphones assume a crisp and sharp voice is highly desirable and that sound frequently fails ACX. It’s a chasing tail thing.

– Select the whole clip or show by clicking just above MUTE.
– Effect > Normalize: [X]Remove DC, [X]Normalize to -3.2 > OK

– Select the whole clip or show by clicking just above MUTE. (just to make sure)
– Analyze > ACX Check.

Screen capture the analysis panel, or write down the top three readings, Peak, RMS (loudness) and Noise. We only need the dB readings. I assume it’s not going to pass.


Write back wherever you get stuck. I plow on until someone stops me.


English version of the ACX readings.

Peak. Look at the blue waves on the timeline. The very tips of the waves are never allowed to go all the way up or down. That’s overload (1.0 on the timeline – 0dB on the bouncing sound meters). That can cause audible crunching and distortion. ACX doesn’t even want you to get close, so that’s their -3dB (70%) limit.

RMS. Loudness. Root Mean Square is a standard electrical measurement which, when applied to audio, happens to work out to loudness. RMS should be between -18dB and -23 dB.

Noise. How loud is the show when you stop talking? This can be a combination of dogs barking next door, air conditioner noise, and the ffffffff noise the microphone is making. All of it lumped into one number. Noise should be quieter than -60dB.

And yes, it’s perfectly possible to have a performance that won’t meet all three at the same time. It’s pretty common for home readers.

It’s also possible to pass ACX-Check and still fail acceptance. If you got there by heavily processing and beating your voice to death, the work will fail Human Quality Control. The voice is distorted and the failure is “Overprocessing.”


Or is there anything else I can do?

Did we lose you?

Many people want us to tell them how to push a button and Everything will Be OK. Not so far. Everybody’s voice, environment and microphones are different, but you all have to fit into the same sound standards. It’s very much a “stop by the store for a fitting.”

You can also record a test sound clip and post that.


I just put up a video on this issue, or at least part of this issue. It may help. If you want to resolve this issue, we can hook up via the www and I can capture your screen and walk you through the process. You can view the video here. Scrub to 9 min and then listen. It won’t cost you a dime.

I just tried to watch the video, but it’s gone. :confused:

The videos are being improved for your viewing pleasure.