ACX loudness requirements: RMS, peak level and noise floor explained

ACX loudness requirements look pretty simple when you first see them.

Your audiobook needs:

• RMS between -23 dB and -18 dB
• peak levels no higher than -3 dB
• a noise floor no higher than -60 dB RMS

That's basically it.

The annoying bit is getting all three to happen at the same time.

Turn a quiet recording up to fix the RMS and suddenly your peaks can be too high. Turn it up again and the background noise comes with it. Push compression or noise reduction too hard and the numbers might technically look better, but the voice can start sounding processed.

So rather than just memorising three ACX numbers, it helps to understand what each one is actually measuring, and how they affect each other.

The ACX audio requirements

For each submitted audio file, ACX currently requires:

ACX also expects the audiobook to sound consistent from file to file. That includes level, tone, noise, spacing and pronunciation.

So getting Chapter 1 through the technical check isn't the whole job. Chapter 12 shouldn't suddenly sound like it was recorded on a different microphone in another room.

For the full official specification, see the ACX audio submission requirements.

RMS is usually the confusing one

RMS is a way of measuring the average level of the audio over time.

Peak level tells you about the loudest moment. RMS tells you much more about where the recording sits overall.

Imagine a narrator reading fairly quietly for a minute and then suddenly shouting one line.

That shout might create a peak of -3.5 dB, so the peak is fine.

But if everything around it is very quiet, the chapter could still sit at -27 dB RMS.

One measurement passes. The other doesn't.

For ACX, the RMS of each submitted file needs to land between -23 dB and -18 dB.

In practice, I wouldn't deliberately aim right at either edge. Something around the middle of the permitted range gives you a little more room either way.

That doesn't mean ACX requires exactly -20 dB RMS. It doesn't. It's just a sensible place to end up.

For the more technical difference between these measurements, see RMS, LUFS and peak.

Peak level is a ceiling, not a loudness target

ACX requires peak values no higher than -3 dB.

A peak is simply the loudest instant in the recording. A hard consonant, laugh, shout or particularly enthusiastic bit of dialogue might create it.

The important distinction is that making your highest peak land at -3 dB does not mean the audiobook is now loud enough.

You could take a very quiet recording, raise it until its loudest point reaches -3 dB and still end up with an RMS measurement below -23 dB.

This is where dynamics processing and limiting can become useful. They can reduce unusually high peaks so the general narration level can be brought up without the loudest moments breaking the peak limit.

But go too far and you'll hear it.

Audiobook narration shouldn't sound crushed just because the meter looks tidy.

Here's where the numbers start fighting each other

Take a deliberately simple example.

You've finished editing a chapter and it measures:

RMS: -27 dB
Peak: -8 dB
Noise floor: -66 dB RMS

The file is too quiet for ACX.

So, in very simplified terms, imagine adding 7 dB of gain.

You'd now be around:

RMS: -20 dB ✅
Peak: -1 dB ❌
Noise floor: -59 dB RMS ❌

You fixed the RMS.

You also pushed the peak above ACX's limit and raised the background noise.

Straight gain doesn't know which bit of the recording is your voice and which bit is the computer fan behind it. Everything comes up.

This is the part that's easy to miss when you're new to audiobook post-production. Making something "louder" isn't one isolated setting.

Noise floor matters more once you turn the recording up

ACX requires a noise floor no higher than -60 dB RMS.

Noise floor is basically what's left when you're not speaking.

Room ambience. A laptop fan. Traffic outside. Mic and preamp noise. Air conditioning. All the boring stuff you only start noticing once you put headphones on.

And if you raise the overall gain of the recording, you raise that too.

A recording can therefore seem quiet and clean while you're tracking, then reveal a much more obvious background once you've brought the narration up to its final level.

This is why a clean recording at source matters so much.

Noise reduction can help. It shouldn't have to perform a miracle.

If you want the deeper explanation of this, see noise floor and room tone.

ACX uses RMS, not a podcast LUFS target

This is another easy one to mix up if you also make podcasts, YouTube videos or other spoken-word content.

ACX specifies RMS for audiobook loudness.

Podcast and streaming workflows are much more likely to talk about LUFS, which is a different way of measuring loudness and is designed to better reflect perceived volume.

They're answering a similar question, but they are not interchangeable measurements.

So don't take a general podcast preset designed around a LUFS target and assume it will also meet ACX's RMS requirement.

It might sound fine.

It can still be wrong for ACX.

This is also why Optivox separates General processing from Audiobook processing. They are being mastered for different delivery standards.

Don't master to the numbers at the expense of the voice

This bit matters more than any exact plugin setting.

Passing the technical numbers doesn't automatically mean you've made a good audiobook.

You could hit the RMS range, stay beneath the peak limit and get the noise floor low enough, while still making the narration sound unnatural through aggressive compression, gating or noise reduction.

You can normally hear when this happens.

Room tone disappears and reappears between sentences. Breaths get chopped strangely. Noise reduction starts leaving metallic artefacts. The performance loses its dynamics.

The goal isn't to make three numbers turn green.

It's to get a natural sounding recording that also sits inside ACX's requirements.

ACX itself also requires consistent overall sound and warns against distracting extra sounds such as plosives, microphone pops, mouse clicks, excessive mouth noise and outtakes.

Technical compliance is the floor, not the finish line.

The order you do things in matters

I generally think about audiobook post-production in roughly this order:

Record → edit → clean → master → check → export

Get the editorial work done first.

Remove mistakes, pickups and unwanted sections. Make sure the performance and timing are how you want them.

Then handle the technical cleanup and final loudness/mastering.

There's not much point perfectly mastering a chapter and then making another twenty edits to it afterwards.

ACX's own production guidance also treats mastering as the final stage after recording and editing.

If you're still working on the recording side itself, see our audiobook recording setup guide.

What does mastering for ACX actually involve?

There's no single processing chain that works perfectly for every voice, mic and room.

That's worth saying because copying someone else's exact compressor or limiter settings from YouTube can easily make your own recording worse.

A typical audiobook finishing chain might involve some combination of:

• cleaning unwanted background noise
• removing clicks and distracting recording artefacts
• EQ where necessary
• controlling unusually high peaks
• balancing dynamics
• adjusting the overall level
• limiting the final output
• checking RMS, peak and noise-floor measurements
• exporting to the required format

The cleaner and more consistent the recording going in, the less dramatic most of this needs to be.

Once you've found a chain that works for your voice and recording space, the whole thing gets a lot easier. You can save the settings and stop rebuilding the process chapter by chapter.

Getting to that point is usually the fiddly part.

Check the finished file before you submit it

Even if you're confident in your processing, check the actual file you're going to upload.

ACX provides Audio Lab, which lets you analyse finished audiobook files against a number of its submission requirements.

One important detail: ACX currently says Audio Lab checks things such as RMS, peak levels, bitrate, bitrate method, sample rate, mixed channels, duplicate files and spacing.

It does not check noise floor or editing errors.

So don't treat one green automated report as permission to skip listening to the file.

Do both.

Check the measurements, then actually listen.

The final MP3 is what matters, not what the WAV looked like three processing steps earlier.

Making ACX loudness automatic with Optivox

This is why Optivox has a dedicated Audiobook mode rather than treating every voice recording the same way.

General spoken-word audio and audiobooks don't use the same delivery targets.

With Audiobook mode selected, Optivox handles the technical cleanup and masters the recording around ACX's audiobook loudness requirements, including its RMS range and peak limit.

The idea is that you shouldn't have to learn a whole chain of noise reduction, dynamics processing, normalisation and limiting just to get a chapter technically ready.

It still isn't magic.

If the original recording has a lawnmower next to the microphone, software isn't going to turn it into a perfect studio recording. And we'd still recommend checking the finished file against ACX's tools and doing a proper listening pass before submission.

But with a reasonably clean source recording, Optivox can take a lot of the repetitive technical work off your hands.

New accounts currently get 60 free minutes, so you can run a chapter through Audiobook mode and compare it with your current workflow.

The short version

For ACX, remember:

RMS: -23 dB to -18 dB

Peak level: no higher than -3 dB

Noise floor: no higher than -60 dB RMS

But don't treat those as three completely separate settings.

Raising the overall level also raises your peaks and your background noise. Controlling peaks too aggressively can flatten the performance. Heavy noise reduction might lower the noise floor while making the narration itself sound worse.

That's the bit a simple checklist doesn't really explain.

ACX compliance is a technical standard.

The audiobook still has to sound like somebody telling you a story.