Sound is way off


Preface: total newbie.

Intent: Produce quality sounding audiobooks for ACX.

Issue: Meeting ACX required RMS and Peak Values destroys sound quality. Comes across as way too loud, robotic, and mechanical.

Settings: Max playback and recording volume. Max speaker and microphone volume. Gain, 0.

An audio recording without any added effects has a very low RMS and Peak value.

Peak level = -15.53 dB
RMS level = -37.33 dB
Noisefloor = -60.64 dB

Followed standard instructions to meet ACX criteria i.e.: effects low-roll-off speech EQ, RMS normalization to -20 dB, limiter to 3.5 soft limit, and noise reduction (based on the noise profile for the first 5 seconds of silence).

The revised audio file meets the ACX check, and the Audiolab verification tool on ACX, but the audio sounds terrible. It sounds like the volume is cranked way too high, with a mechanical tonality to it.

The ACX check doesn’t always work. In order to meet the Audiolab verification too, I’ve had to program audacity to the following settings:
-Peak level: cut off to -6.5 dB
-RMS normalization to -23
-Noisefloor to -25

Any advice would be much appreciated.


Check Windows “audio enhancements” are disabled
Disable recording & playback enhancements for faithful sound.

BTW there can also be a second layer of “audio enhancements”
such as MaxxAudio, or DTS, (depending on computer manufacturer).

That’s very low. Here are some possible explanations …
Speaking into the insensitive side of microphone.
Accidentally recording from computer’s built-in in microphone.
Not supplying phantom power to a mic which requires it.
Mic level slider in Audacity not turned up to maximum.
Mic has gain knob which is not turned up high enough.
The mic level slider(s) within Windows sound not turned up high enough …


Thanks so much for the reply. I really appreciate you breaking it down. I disabled the “audio enhancements” which improved the quality a bit. The RMS level is still way too low, i.e.: in the -30 range. I’ve gone through the settings, and none of the explanations you’ve highlighted apply. I’m using the MAONO AU-PM421 Mic, which is a budget microphone, however the manufacturer does not identify the requirement for phantom power.

Could it be anything else…?


It’s a USB mic, so does not require phantom power.
It has a gain knob on it, that should make things louder/quieter …

Mic Gain Knob: Flexible Gain Knob can increase the volume and sensitivity of the USB gaming microphone,

The mic gain knob must be facing the performer: that’s the sensitive side of the mic …


Ah. The sensitive side wasn’t quite facing me. Having adjusted my mic, the RMS is much better hovering around -20. Sorry for the rookie mistake, and thank you so much for taking the time to troubleshoot it with me.


Exactly correct. But there’s way to not work that hard. We publish Audiobook Mastering which is a one-step tool to automatically apply all three effects.

There is one soft correction. You should not be applying Noise Reduction customized for each chapter. ACX demands all your chapters match. If you do require Noise Reduction (some recording processes don’t) you might see if Noise Reduction of the Beast (6, 6, 6) will do it for you. That’s a nice, gentle reduction and nobody can tell what you did. The fuzzy rule is passing the -60dB noise limit by at least 5. So -65dB.

Unless they changed it, on-line ACX Audio Lab doesn’t check noise. Audacity ACX-Check does.

It’s good to know what your noise is. If it’s gentle, Spring Rain in the Trees (shshshshsh) then that’s normal, although a USB microphone should not fail that right out of the gate.

There are other noises, some of them pure evil. There is a USB error caused by both a cheap, inexpensive computer that doesn’t cost very much and a similar microphone. It causes digital trash to leak between the wires of the USB cable. It sounds like screaming mosquitoes and more importantly, will cause your voice submission to fail Human Testing even though the file passes ACX-Check.


If you’re having distortion problems you may have some odd errors. Post a sound sample so we an hear it. Nothing like trying to guess at a mystery problem across multiple time zones. Sound samples help.

Don’t help it. Just post what you got.



I’m not sure what I’m doing wrong.

Setup: As soundproofed as I can make it, the noise floor is no longer an issue. It is often still within an acceptable dB range before applying noise reduction. I’m the suggested length away from the microphone, with max recording/playback volume. All audio enhancements on my computer have been disabled.

Issue: I can’t seem to get that crisp, precise, yet human tone, see sample.

As you can tell, it sounds throaty, monotone and harsh.

To be clear, with the sample I provided - what I did in sequence was noise reduction via 6,6,6 and apply the Audacity Masterclass Plugin. That’s all.

Now when I reverse that order, i.e.: Audacity Masterclass Plugin then noise reduction (6,6,6,) it sounds loud, and harsh.

The obvious question is whether I actually talk like that, so I’ve attached the same recording without effects so you can judge this.

In my opinion - it sounds (improvement needed for siblance/whistling - got it), much better than when I start adding effects/noise reduction, and I’m trying to figure out why.

Now… if I give the track a ‘bass boost’ it restores the human element, making the track feel more warm, but discernably equalized? If I can say that.

What would you recommend? And as much as what you would recommend, what’s the sequence in which you would action it? I’m reading a lot of conflicting information where some people start with noise reduction, then compression - whereas others use EQs, then noise reduction. I’ve played around with various combinations, and they all sound awful to be frank, hence my reaching out to you, yet again.

Hope you can help.


“Not to give credit…”

That may not be the best test. It’s the beginning of a parenthetical phrase and it’s really short. It sounds like it should be followed by an unspoken comma and then the real declarative sentence. All that and it’s an MP3 file.

Never do production in MP3. It creates distortions. Use perfect quality WAV for all your file exports. The first and only time you should hit an MP3 is when you create that special one for ACX. And then once you make it, you can’t ever change it. Your edit master file should be a WAV.

Use the format and words I posted for a forum test.

Don’t forget the hold-your-breath silence at the beginning.


ACX-Check and the associated on-line ACX Audio Lab only check the technical specifications of your file.

Is it a file?..Yes.
Is it loud enough?..Yes


Your complaint is about the theatrical presentation of the work. Does it sound pleasant? Is the reading rhythm about right? Etc. There’s no good test for that which is why I requested the sound file.

And don’t post the fancy ones. Just the simple, clean reading. No effects.



First, thank you so much for taking the time.

As req’d, here is the sample.

Looking forward to any comments you have.


There is a tick or pop at about the 2-1/2 second mark. Do you know what that was? Did you turn something on or off?

I think I identified one problem right away. You have a very relaxed presenting style. Past the theatrical problems, you’re not getting loud enough to overwhelm the microphone’s natural noise. I can force the last file to pass ACX, but I have to apply very strong noise reduction to get there. Strong noise reduction causes tonal distortions such as essing and other sharp sibilants. I think you noticed this already.

Your submitted file is almost perfect in volume, so we may need Oblique Positioning.

Get closer to the microphone. But you can’t just skooch it closer, that will give P-Popping and other mouth distortions.

Try pushing it off to one side while still aimed at you (B).

You should be able to get it about one fist away from your face and significantly louder. You may need to turn down the microphone gain control to avoid overload distortion, but that will reduce the noise (we hope).

And now for the theatrical problems. The pause between “Rich” and “Creamy” is in the wrong place. It’s not where a natural pause would be. It’s a trick script. It’s a little too long to take in one breath, so most people have to put a breath somewhere.

Read it like a radio commercial. It’s not at all unusual for announcers to have slightly different voices depending on content.

Have you ever heard reading aloud to children at the public library? Arrange to be in the audience the next time one of those goes by. Then compare it to your style.



Thank you! I’ll try that. The pop 2 1/2 seconds in was just me clearing my throat… normally I would delete that, but you asked for a raw (unedited) version, so I didn’t bother.

How much do you think I should reduce the gain???

And on a scale of 1- 10, 0 being abysmal, 10 being professional radio anchor; where am I situated? Frank feedback would be most appreciated.


Exactly correct. However, it does throw us off when sounds don’t match. Stop me if you’ve heard this, but there was a recent forum poster whose natural voice sounded like a broken microphone.

It’s nice to think the I can give you a number or a position of the knob, but really, you should speak in your normal presenting style and record so that the Audacity bouncing sound meter occasionally pops up to about -10dB. That’s your presentation setting. The blue waves can peak up to about half-way.

I don’t see that microphone has a place to plug in headphones. Live monitoring your own voice is highly recommended when you have to watch the script and the sound meter at the same time.

This is a stand-alone sound recorder, but you get the idea.

You would think you could just plug your headphones into the computer, but those connections and wireless earphones can have distortion, echo, and delay problems.

I knew a pro radio presenter where I used to live and he was a shock the first time I met him. His natural voice and speaking style was just terrible—almost broken. But you put him in front of a microphone and it was like someone threw a switch to top-quality broadcast announcer.

So that can be done, but be advised he wasn’t announcing long-form audiobooks. About the longest work he ever had to present was hourly local news, occasional weather emergency, and introduce the network news on the hour.

Once we get basic voice recording nailed, ACX has a service called 15 Minute Checkpoint.

I haven’t read down through the whole thing, but they used to offer a long-form clip quality evaluation—and then they stopped because of the pandemic. Now they have this.

You may be able to record a piece, submit, and get their opinion of your viability.


Found it.

This is roughly what normal looks like.

I know you’re going to say, "Every time I get the computer close enough to see the meters, the fan noise gets in the show.

Yes. It does.


This topic was automatically closed after 30 days. New replies are no longer allowed.