AudioBook Mastering version 4

Sounds like you need a De-esser plugin …
There are two free ones made specifically for Audacity …
Paul-L’s de-esser precise, but complex & can be slow.
Steve’s de-esser plugin less precise, but simple & fast.

Your RMS is your overall loudness. So when your RMS is too quiet, the fix is to make ALL the audio louder. This probably messes with your peak (no big deal, the limiter step in audiobook mastery solves this perfectly). Your noise floor, on the other hand, is not an easy thing to fix. Generally speaking, if RMS normalize makes your noise floor too loud, you can either try to make your recording space quieter or apply very gentle noise reduction, -6 -6 -6, the number of the beast, can work. But if you start getting past -10, you’re probably going to end up with some damage on your audio.

Which means the best way to handle the situation is usually looking at your recording area and figuring out what you can do to make it quieter, or make -you- louder. Getting closer to the mic, creating a sound isolation area for your mic, cozying up in a closet, all valid ways of doing it. There are many, many more ways too.

But when the file is ACX Passed, the audio doesn’t sound good.

Both have to pass. ACX Check should get you past the ACX “Robot”, the hardware test, but then you have to sound good too. That’s ACX Human Quality Control. That’s where you die if you needed intense filters, effects or corrections to pass ACX Check. Anybody can sound like a bad cellphone or a broken microphone. The ACX goal is listening to someone tell you a story in real life over cups of tea.

I just need to produce an Audiobook, so I am desperate to get it to pass ACX.

Write a check to a professional studio, walk in and produce the sound files. Even better if they have a reputation for successful audiobook production. You may be able to get that as part of the contract.

Home microphone makers insist on advertising that you can set up on your kitchen table and crank out book after book. Maybe not. I have a soundproofed third bedroom (one of the kids of a former owner played drums) and even with that I have to wait until most of the street traffic goes away at night. This is not easy.

The longest forum message is 39 chapters and over a year. The poster just wanted to read audiobooks in his apartment in Hollywood. We did eventually get him working. He’s the champion.

We can do this as long as you can, but remember, once you get it working, you have to keep it working through a whole book. ACX likes all the chapters and segments to match.

This is where you Google studios in your area. You might be able to meet in the middle and hire-on someone to record you where you are. I know people who used to do that.


Hi, I went through all the steps as you suggest, and got the first wo passes but then when I adjusted the noisefloor ( which was only 0.2 dB out -59.8dB) the RMS went out by about 2.8db, then I re did that and the others went out again. How did I get on this merry go round?

You’re failing noise. That’s very common for home readers.

There is another way to look at this. When you announce, your noise has to be 1000 times quieter than your voice. That’s the -60dB number in English. That’s including noises the computer and microphone are making.

Passing noise is hard to do and if you miss it, you get in that juggling act of only being able to meet two of the three Audiobook numbers. “If I make my voice louder, then the noise gets louder…”

Record and post a 20 second WAV sample of your work.

Go down the blue links. They’re short and they give you good hints on how to do each step. Don’t adjust anything or apply any effects.


Ok, just upgraded to 2.3.3 and went to apply this technique to a podcast I’m working on and turns out Equalization is now replace with Filter Curve. I don’t see the settings we used, rolloff for speech, in there. Any advice to this and the other settings described here for 2.3.3?

Until we sort this, here’s the curve you should be able to install in the current equalizer.

LF_rolloff_for_speech.xml (299 Bytes)
"LF Rolloff is the curve’s old name. It’s the same curve.


“Low rolloff for speech” is already included in the “Filter Curve” effect.

In the “Filter Curve” effect, click the “Manage” button,
then in the “Factory presets”, select “Low rolloff for speech”.

When I go to Effect->Add/Remove Plug-ins I get a new page that shows them, with controls to enable/disable but no apparent way to “add” this.

You seemed to allude that I could install this in the new Filter Curve. Don’t see that either. Although I do see under the new filter curve there’s a Manage button and then Factory Presets and one choice is Low Rolloff for Speech. Don’t see how to set it to 5,000 though.

They can’t be enabled because they’re not actually there. Unfortunately the Effect Manager does not yet clean out old effects that are no longer present, so the old “Equalization” effect is still listed, even though it no longer exists.

You don’t need to.
The old Equalization effect used quite a small “Size” setting by default, hence the advice to increase the Size setting. The new “Filter Curve” effect already has a larger “Size” set internally (not user adjustable), so you no longer need to do that.

Dumb question, but I got the 2.3.3 update today and was wondering how to adjust the filter length for the “Filter Curve” (or Equalisation). You used to be able to adjust it, was it scrapped, or am I just blind? I’m having to re-do my Macro (or ‘Chain’) and going through the fiddley bits.

I was wondering when that was going to pop up.

The old Equalizer had quality issues. When you used it, it would present with two different curves.

The blue one you (or the preset) drew operated in the subjunctive. This would have been the effect you might have gotten had the tool followed our desire. The green curve was the one you actually got. It was a little sloppier.

The new tool, Effect > Filter Curve has no such error, or the error is smaller and better behaved. So there is no adjustment.

One critical feature of the effect is that it should have good suppression at 60Hz which is the power frequency in the US. This can help suppress microphone hum problems.

The 50Hz power (lower pitch) in the UK is easier to deal with. If you start the curve much higher pitch than about 100Hz, you can start to hear it working in the voice or performance.

It’s a juggling act.

This effect appears on many field sound mixers to help with popping and rumble cause by wind.


Thanks Kozikowski for the explanation. I’m not an audio tech-head, but this is good stuff to know. I’m really glad that I saved picture references of my old EQ specs. I do miss seeing the graph with the graphic EQ, it feels weird without it somehow.

I’m lost.

That’s a screen capture of Low Rolloff under Effect > Filter Curve in Audacity 2.3.3.

Screen Shot 2019-12-09 at 18.36.32.png
You can open the Low Rolloff in any text editor and transfer the points manually by dragging the curve around with your mouse.

Screen Shot 2019-12-09 at 18.39.01.png
For one example, the last point in the list is f=100 d=0.0. In English, voice tones at 100Hz (ballsy announcer low tones) are not affected by this filter. See the tiny circle in the green line above at the cross of 0dB and 100Hz?

Watch. I’ll grab it with my mouse and mess it up.

Screen Shot 2019-12-09 at 18.45.09.png
Its a pretty simple curve for all the tricks it does.



This is my first post to this forum. Thanks for making it available. Please excuse in advance my naivete and asking questions which may appear stupid.

This is my first attempt at making an audiobook. I use windows 10 and Audacity 2.3.3

I have downloaded all the plug-ins I am aware exist. You referred to an ‘equalization’ effect I can’t find anywhere. Is this the same as normalization?

When I manage to get peak values below -3dB the RMS shoots above 23dB and when I try to get the RMS below 23dB, peak values rocket above -3dB. How do I conciliate both? Does a plug-in exist to achieve that?

I have been doing this for several hours tweaking figures without any success.

Many thanks in advance for your help.

It’s now called “Filter Curve” and it’s now become easier to use (no need to set a “Size” parameter - that’s now automatic).

In the next version of Audacity it will probably be called “Filter Curve Eq” (or something similar).

When I manage to get peak values below -3dB the RMS shoots above 23dB

There is an audiobook suite of tools that can deal with all that.

That’s the detailed step by step. This is the quick version.

Screen Shot 2020-01-07 at 16.45.44.png
If your original voice recording is loud enough and you recorded in a very quiet room, you may be done.

But back here on Earth; that never works. Home recordings never pass noise. Between the computer fan noise, refrigerator, (fish tank in one case) and the metrobus outside, noise is a constant struggle for home performers. The ACX Audiobook specification for background noise is -60dB. In English, that means your background noise has to be 1000 times quieter than your speaking voice. That’s not a typo.

Record a test (if you’re set up enough) and I’ll run it through mastering and see how it goes.

Read down the blue links. They’re very short.


dB or Decibel Relationships

<— Quieter ———————— Louder —>
-65dB, -60dB, -23dB, -18dB, -3dB, 0dB

That’s all the audiobook dB specifications.

— Maximum volume is 0dB. Sound gets quieter with negative numbers.
— -3dB is the loudest audiobook peak sound value.
— RMS or loudness needs to hit between -18dB (louder) and -23dB (quieter).
— Noise value has to be quieter than -60dB.

Noise is the troublesome one. Actual submitted performance background noise should be quieter than -65dB in real life. There is no submitting performance noise of -60.2dB, for example. Technically, that does pass, but sooner or later, one of your audiobook chapters is going to fail Noise and you will need to apply noise reduction or some other fix. That fix may make one chapter not match the others and that violates the ACX law that all your chapters match.

It’s possible to use dB for measurements other than the above. You can use them in comparisons. “My voice recording is 3dB louder than yours.”

Sound Pressure Level measures sound volume in free air with special meters. DB SPL uses positive numbers. “My truck engine sound measures 125dB SPL” (threshold of pain). There is no direct relationship between dB (audiobook, above) and dB SPL. Some Phones claim to be able to measure dB SPL directly.


Hallo vielmals,
finde in der Version 4.2.1. Roloff of speach nicht. Den Equalizer, der diese Funktion in den älteren Versionen beinhaltet hat, lässt sich nicht aktivieren. Benutze jetzt Filtercurve. Ist das in Ordnung?
Vielen Dank für Eure Antworten.


I’m attaching the test file as described above.

I can see a problem right away. My peaks only reach about -25. I have the mic cranked to 100 both in Audacity and in the laptop’s control panel. I’ve also tried using a different USB port, without much improvement. I’m using an aged Audio-Technica ATR2100 microphone. Some Googling has revealed that low recording levels are a problem with this particular model, but I used to record podcasts and video voiceovers way back when with this mic and I never encountered this problem before. Though mic placement might be a factor. I used to hand-hold the mic right in front of my mouth before. Now I have the mic suspended a few inches from my mouth, as advised by the ACX audiobook recording tutorials.

My attempts to fix the low level in Audacity using the Filter Curve, RMS Normalize and Limiter make the noise floor way too high. See attached screen shot.

Any ideas?

Thank you,

ACX Check test.png