Audacity ACX Check plug-in vs. ACX Audio Lab

I noticed that when I run the ACX Check plug-in, the results are slightly different–sometimes significantly different–than those with ACX’s Audio Lab tool.

Could this be because I don’t have the latest version of the Audacity plug-in?

The “ACX Audio Lab” tool is a simplified version of the robot that ACX use for the initial assessment of audiobook submissions.
I’ve not found any technical details about how exactly it works or what precisely it does.

The “ACX Check” is an Audacity plug-in that we created to give measurements based on the specifications published by ACX (
I can give full technical details of what this does and how it does it because it is “open source” and the code is freely available (also, I wrote the latest version :wink:)

Basically, they are both doing similar things: checking if the audio meets the published ACX specification, but they probably do so in slightly different ways.

The “ACX Audio Lab” tool has some additional checks regarding the file format.
The “ACX Check” plug-in also does a basic test for noise floor.

If your audio fits towards the middle of each specified range, it should pass both tests.
If it is on the edge of one of the ranges, then that could be a problem.

The latest version of ACX check is here:

the results are slightly different–sometimes significantly different–than those with ACX’s Audio Lab tool.

Do you pass both?

We should note that surgically inspecting Peaks and RMS (Loudness) isn’t as important as you think. Home Readers never pass Noise. The Noise specification is -60dB. In English, your room noise has to be a thousand times quieter than your voice. And that’s just squeaking past. Our recommendation is -65dB and quieter than that is even better.

Stop what you’re doing right this second and listen. Can you tell if your computer is on just by listening? Kiss of death. I got traffic noises, the refrigerator chillin’, and two escapement clocks without even trying very hard. If I try really hard, I can hear jets going over.

How did you do?


ACX has a pre-technology test.

Are you reading a book I can buy on Amazon in any form right this second? That’s required.

Also, don’t read anything on this list (scroll down).

No, they won’t let you read a cookbook.

There is one production requirement that a surprising number of people fail. You don’t just open up and start reading. There is a required amount of room tone (gentle room silence) at the beginning and end of each chapter.

Screen Shot 2021-06-10 at 3.42.18 PM.png

Trying to understand here… so if I fail the noise floor on ACX check but pass with No Issues Found on audiolab then can I submit a final MP3 or not?

Please be more specific. What were the actual results from ACX Check and AudioLab?

ACX Check is not the only way to check noise at home. The legacy method is by carefully selecting the area in your performance that you consider Room Tone or background noise. Another way to think about that is what you record when you stop talking and hold your breath for a second or two.

Then Analyze > Contrast > Foreground > Measure Selection. That’s the RMS (loudness) value of the noise in your selection.

We want to measure noise because it’s important to ACX for submission and they said so.

Screen Shot 2021-07-28 at 7.19.12 AM.png
I haven’t sat through AudioLab testing in a while and they may have changed it, but even though they do not check noise, they do say you have to pass all the technical specifications listed, and noise is in that list.

Traditionally, Both ACX Check and AudioaLab are first-pass testing. After you make it through them (hopefully, both), you advance to Human Quality Control where an actual person listens to the work for theatrical quality. That’s the one I failed. I submitted a technically perfect test in every possible way, but they didn’t like my tongue-ticks, lip smacks, glottal emphasis, and other mouth noises. Basically, my voice sucked.

So you should keep that in mind before you bet the ranch based on any of these tests.

ACX made it painfully difficult for performers, particularly new ones, to succeed. They used to do a full acceptance analysis, humans and all, and offer personal guidance for success. That’s what happened to me.

Thanks for submitting your sample, “MaxwellsCafe3.mp3”, for review. You are off to a good start here but there are improvements needed before we’d consider you ready for audiobook production. Below are the issues to address:
. . .
The ACX Team

Not anymore. They offer ACX AudioLab and the only way to submit is to read the whole book.

The forum can do a “real” quality control analysis for a short test.

It’s ten seconds. Not too painful.


results are slightly different–sometimes significantly different–than those with ACX’s Audio Lab tool.

There are differences of opinion about how to measure some of those values. For one example, do you measure the loudness of dramatic pauses in reading? They have no loudness and a technique difference here will affect the overall loudness of the measurement. I believe ACX ignores gaps and Audacity doesn’t.

When you post notes like that, be specific about what happened. We can’t analyze an open-ended statement.

The unhelpful but accurate answer to that note is, “Yes.”


Another thing that flummox’s new users is dB values. They appear backwards because they’re negative numbers.

This is why I stopped calling sound values higher and lower and started using louder and quieter.

There are a couple of systems like that. Performance rhythm timing or “latency” works in negative numbers and it will make you want to go lie down for a while after struggling with it.