RMS Normalize makes audio sound tinny


I have narrated over a dozen books for ACX and only recently discovered the ACX check tool. What’s interesting is it turns out the files I’ve submitted in the past didn’t meet the RMS requirement, but were still approved! Anyways, I’m working on a new book now and have installed the ACX check plugin and the RMS Normalize plug in.

So, here’s the problem: When I run the RMS Normalize plug in, it makes the audio sound tinny, like I’m in a metal box. For comparison sake, I tried different effects (Normalize, Amplify, Compression, Limiter) separately or in a sequence and I don’t hear that tinniness. My raw file is super clean edit, recorded in a very quiet room. I’m running the Low Roll Off for Speech EQ before the RMS Normalize. Audio sounds fine after EQ, but tinny after RMS Normalize.

Has anyone else had this problem?

I tried Koz’s sequence" AudioBook Processing Tool Notes found here: http://www.kozco.com/tech/audacity/audacity.html but the result is just shy of the RMS needed to pass ACX.

Suggestions? Thanks.


Has anyone else had this problem?

You’re number one!

Something is broken. Normalize and RMS Normalize are cousins. They both change the volume of a clip and nothing else (Normalize can change DC level which is slippery because DC is a microphone defect and not sound).

Normalize, in common with most tools in this group, does it by sensing peaks or tips of the blue waves. RMS Normalize, oddly enough, does it by sensing RMS or Volume. RMS and Volume don’t track exactly. You can get into religious wars with defining volume, but RMS is a very specific electrical measurement, so it won.

Steve designed RMS Normalize to meet the RMS requirement in ACX. Until then, you had to get there by running around the barn and guesswork with other tools. You should probably trash your existing RMS Normalze and pull down a fresh copy.

Are you following Audiobook Mastering 4?

You are warned against mixing and matching effects in mastering. Mastering 4 may not look like much, it’s only three tools, but they clean up after each other. The addition of compression, for one notable tool, can mean some of the cleanup won’t get done, or it can create more damage than it clears.

Also be advised that you can fake out ACX Check. ACX Check needs at least 3/4 second of clean, pure Room Tone (background noise) to accurately measure Noise. Most natural speech recordings don’t have that.

It’s super critical to tell us what exactly failed. “It doesn’t pass ACX Check” doesn’t tell us enough. Mastering 4 will always give Peak and RMS exactly correct and the only variable is noise. There is no “It doesn’t pass RMS.”

What process did you use for the earlier books? I’ve said multiple times if you have a successful process, no matter who designed it, you should hold onto it with white knuckles.

Are you processing stereo voice, two blue waves? Did you get a better microphone/computer/interface? I see you’re using WinXP. Which Audacity are you using—all three numbers? You should probably be using 2.3.0.



If RMS Normalize is making a big change in the volume, subjective perception of its equalization will change :
as the perceived frequency-content is volume-dependent … https ://en.wikipedia.org/wiki/Fletcher–Munson curves

A psychoacoustic phenomenon.

On the other hand, if you’re recording in stereo, two blue waves, there are ways of connecting a microphone incorrectly that can produce very bad effects with RMS Normalize as it manages left and right independently.

And no-one wrote you can’t have more than one problem.


RMS equalization alone can push the peaks over 0dB and you can get [u]clipping[/u] (distortion). In fact, you are likely to get clipping if you normalize/adjust the RMS levels to meet ACX specs and you don’t do anything else. (And it won’t pass ACX if it’s clipped.)

But if you don’t push the levels into clipping, normalization is simply a volume adjustment and it doesn’t affect the character/quality of the sound.


And this is where you post and tell us more about what you’re doing.