Non-Matching ACX Checks export into .mp3

Using Audacity 2.3.2 on Windows 10 Enterprise

I am exporting a file as an .mp3 The original passes the ACX Check (Peak -3.01; RMS -22.8; Noise -69) . That exported file fails when uploaded into ACX Audio Lab (although it doesn’t Flag – “No Issues Found”; the downloaded parameters show hot levels) as (Peak -2.82; RMS -23.12; Noise -76.42 [Pass])

When I Re-Import the mp3 file into Audacity and re-run ACX Check on the import, the new ACX Check matches the Audio Lab results.

The book I am doing is a Trivia – as such, I am using a lot of Sound FX ( Syncopated Clock & Transitional Swooshes). I am mixing down to a single, stereo track before the export. I’ve checked the Gain (0db) and the L-R Channel (Center). Remixes to a single channel results the same. I’ve tried exporting with the gain (-1 db), but then the Peak matches the lower setting, but tends to really mess up the RMS (Around -25).

What am I overlooking? How do I adjust?

I read that four times and I can’t find the part where you tell us what’s broken.

Give it another pass.

If you’re planning on submitting to ACX, they tend to not like auxiliary sounds or music. Can you prove you have copyright for all that stuff? We assume you’re submitting for payment?


RE: Copyright: I went to a fee sfx site, and paid the minimal ($3) fee for use under 10,000 sales (including paid projects). I have used other efx/music jingles like this before for a couple of other projects.

What I see as broken: On export to an mp3, the Levels are increasing by .03 - .07 db, RMS decreasing by about the same or more… It doesn’t follow a correlation if I just reduce the levels by the corresponding amount.

I even imported the mp3, did a hard limiter to peak of-4.5db, normalize to -3.4db, reexported it with success. Then tried it with another file (apparently, several of these sfx file failed, but didn’t flag), roughly same levels, and it made it worse.

Retried the same file (thanks, Ctrl+Z!), no changes in levels, re-exported, re-imported into another Audibles file, and it passed Peak, but failed RMS.

Thanks Koz – you da best!

I went to a fee sfx site

You should tell them that if they ask.

MP3 gets its small, convenient, reasonable quality files by rearranging sound tones and leaving some of them out. It’s that last part that makes MP3 dangerous because once you make one, you can’t back up. There are parts of the show just flat missing.

It’s the first part that gives you MP3 files that don’t match the WAVs they were made from. Nobody wrote anywhere that the MP3 has to be exactly the same volume as the original. It can wander a bit. We should remember that MP3’s full family name is MPEG-1, Layer 3. It’s the sound channel for a video file, not a precision production tool.

the Levels are increasing by .03 - .07 db

That’s just scary. The ACX specifications are peaks quieter than -3dB, Noise quieter than -60dB and RMS (Loudness) somewhere between -18dB (loud) and -23dB (quiet).

If you apply the Audacity Audiobook Mastering Suite to your work, Peak will pass, RMS will end up in the middle (-20dB) and if you recorded in a nice quiet room, Noise will pass. I’ve never experienced a WAV or Project that passes Mastering and ACX-Check with reasonable elbow room that then failed making the MP3.

As a first pass, I would say you’re not using Mastering. It’s possible to get to ACX using compression and other magic tools, people have posted that, but I couldn’t do it reliably.

This is the short form.

You should be exporting your Edit Master as a WAV (Microsoft) 16-bit file and only then make the MP3 for submission.

Is any of this a surprise? Note there are no severe hack and burn tools, clippers, hard limiters, or compressors. ACX Mastering was designed to sound exactly like you, except passing.


Please note that Mastering is a “suite.” A harmonious grouping. It’s designed to work from an original reading. Don’t add any tools in the middle or leave any out.

The tools not only work effectively, but they clean up after each other. For example, the first tool, Filter Curve gets rid of sub-sonic home microphone errors, rumble, and thunder which can throw off the other tools. It was designed after Broadcast and Hollywood tools which do the same thing.



I still have to play with the mix – but instead of “Riding the Edge” of the (Peak/RMS) limits, I shoot for more “headroom” --shoot for the -19/-20 dB RMS and instead of a tight -3.01 dB peak, I’ll go -3.1 to -3.5 (a little more headroom for music/sfx projects).

The other is I run the ACX Check ALL THE TIME! after recording, I get the track into the sweet spot before I edit (I keep a “master” track untouched, just in case. This has saved my posterior on multiple occasions).

I also run the check before exporting/conversion to mp3. THEN, after export, I re-import (drag and drop) the mp3 file back into the master aup3 file and run the ACX Check one last time. The results are pretty spot-on what ACX says the results are --the same 80%, and only .1 to .2 off the rest. Its a lot faster than uploading the file to the ACX Playground and waiting… oh the waiting!

I’ll go -3.1 to -3.5 (a little more headroom for music/sfx projects).

Pretty much why the limiter in the Audiobook Mastering Suite is set for -3.5dB.

Its a lot faster than uploading the file to the ACX Playground and waiting… oh the waiting!

ACX Audiolab and ACX Check give very similar readings—we’re both working from the same technical standards, but Audiolab doesn’t check noise. So if you’re heading for an audiobook submission, you’re going to need that.

Home Readers never pass noise. You can get to Peak and RMS with conventional tools, but Noise is the college course.