Soft voice and microphone levels

Hi Folks. I am new to recording and I have read some posts about USB vs Preamp and conventional mic but am seeking some experience between the two. I have an old model AT2020USB that I have been recording with. I am fairly happy with the mic but it has its downsides and I wonder if I can do better. Several of my complaints are how the mic sounds different with only 1mm of movement in any direction and the recording volume I have to crank it up to. I nearly have to have my head clamped down and do constant distance and head angle checks to sound consistent. if I move my head a bit to flip a page, I have to speak a few words while weaving around to get the sweet spot back.
I am using Audacity 2.1.2 on several Windows 7 and higher pc’s and laptops. I can consistently record on each device and with only minor adjustments to the recording, get all levels where I need them to be and so far so good with ACX. But, in order to get close enough with RMS so that I do not have to amplify very much, I have to have the microphone recording level set to at least 97. At 97, even in my ultra quiet sound booth, I can hear a fly pass gas two blocks away, not to mention every spit bubble and click. At 97, it is so sensitive yet I still have to amplify a few db. I have been working on improving my performance in the clicking area and am getting better at that. I have learned how much to open my mouth, how to quiet my esses and how to close down some words and where to keep my tongue etc…

Anyway, I have a fairly deep but soft voice (I almost feel like the duck quack experiment on Mythbusters. It’s like my voice cancels itself out). I am not going to change out my mic in the middle of two books, but am very curious about trying a focusrite scarlett solo and trying different gain settings and mics to get a natural sound that is within %10 of where my RMS needs to be. I imagine that having a separate gain setting would be preferable to the wide open USB interface? I know all mics are different so I guess I am specifically looking for any experience with pre-amps and if that is a preferable setup for softer voices compared to a wide open USB only mic.

Sorry for the long post and my inexperience and thanks for any input.


a fly pass gas two blocks away

Just to cover the basics, you are speaking into the Audio Technica logo, right? The 2020 is a side-fire microphone. You don’t talk into the back (labeled “back”) and you don’t talk into the round top.

You should check with Audio Technica and see if there is a Windows 10 driver for this microphone. I don’t expect one, but it’s good to check.

the mic sounds different with only 1mm of movement in any direction

That screams using the microphone wrong. This is a cardioid (heart-shaped pattern) microphone, top view.
Screen Shot 2016-12-13 at 22.02.22.png
You are intended to stand and speak into the “zero” part of the curve, which is where Josh is singing in that illustration. If you’re announcing into the back, the slightest movement of your head will give you extraordinary quality and volume changes, and you will be able to hear the clock in your car in the garage.

If you are speaking correctly, then you’re just too close. Nobody is giving USB microphones any awards for high volume. Almost all of them run at low volume to keep you out of trouble. Low volume can mostly be made up in post production processing and amplification. Low volume is always assumed to be the performers shortcoming.

High volume produces immediate, severe sound damage and causes you to send the microphone back.

So low volume it is.

Our published goal is for you to hit -6dB (yellow region) of Audacity’s sound meters occasionally during your performance.

Maybe you will not be able to do that with your setup. Can you reliably hit -12 dB or -15 dB? That’s not dreadful and may help with a lot of your extreme sensitivity problems. It’s not unusual to run a USB microphone full up volume.

We are publishing new tools for meeting the ACX technical conformance. They’re much easier to use than the older ones, but I suspect don’t sound the same, so you can’t change over in the middle of a book.

We’re starting to notice that Noise Reduction makes sibilance worse. If you have to use a lot of noise reduction in your performances, it may turn your normally crisp ss sounds into a harsh, piercing sound that will crack eggs.

What is your correction suite? Get done announcing a chapter and then what?

Step 1. Immediately Export a WAV (Microsoft), 16-bit protection copy and move it to a separate drive so you don’t damage it by accident.

Take it from there.


I’m just reading that again. Are you trying to directly read into ACX delivery format? You don’t mention post production, processing or effects in the whole post. I don’t think normal humans can do that, nor do I think it’s required which may be why you’re struggling.


Thanks for your time and efforts Koz!

I guess I have wound up trying to get as close as possible with minimum processing.
The current process:
Record at 97, levels are good there and into the yellow here and there when I am speaking naturally. I am doing the “hang ten” thumb to pinky distance from the mic and am aiming for the blue light inside. I do use a pop filter.
I normalize, run one round of noise reduction, run the de-clicker and de-esser and then check contrast and calculate the adjustment and amplify to get to -22db (usually 1.5 or so), then I run the limiter set at 3.5 hard limit and I’m done and the chapters pass the ACX check and I upload the MP3 to the site. Maybe I am recording a bit too loud?

When I was struggling (guessing) to get any real consistent process to meet the basic ACX check; treating with equalization and having to endlessly amplify to pass, I just kept sounding more and more mechanical, especially as I amplified my voice when I was recording with the level set at 50. I have learned some things since then and at least have a better understanding of the whole process but still strive for improvement on all aspects. Somewhere along the way, I read that you should record within 10% of your desired final level as if you record too low, there is not much there to amplify and it does not sound natural. When I recorded in an untreated closet at 50, the noise floor was better than recording in an audio booth at 97. Obviously, the acoustic quality of the recording is better but the base noise level is worse and my microphone is ultra, ultra sensitive. I have tried to record at lower levels in the booth but when treated to pass the ACX check plug-in, I start sounding like I am hollow and in a can.

I have a lot to learn and appreciate any shared experiences. Just really curious if a pre-amp and different mic would be any different for me and allow me to have a better experience than the cranked up USB setup.


I need to go back and read through that again. If you have a process that works, we generally step back slowly and leave you alone.

The process is not supposed to be pulling-teeth painful. After you get the swing, you just have to pay attention.

We have a new tool coming on-line that should make meeting ACX a little easier.

Are you using ACX-Check?

As we go.


into the yellow here and there when I am speaking naturally.

Exactly correct. That’s the sweet spot between too loud overload and too low noisy.

I am doing the “hang ten” thumb to pinky distance from the mic and am aiming for the blue light inside.

The blue light is new. What do the instructions say? Hang-ten spacing is just about right. Do you use a round pop and blast filter like Josh is using?

run one round of noise reduction,

At what settings? First number much stiffer than 12 starts to sound like talking into a wine glass. I was going to ask if you were using ACX-Check.

What happens if you apply ACX-Check to a raw recording? No processing. Just announce a short segment with your normal presentation and run ACX-Check. Use this format.

Then post it here and we may be able to suggest other processing.


Thanks Koz,

I am using the default noise reduction setting in the stock plugin. I will submit a sample tomorrow. This will really help me, thanks!

I have a much smaller round pop filter than the photo but have also made my own, which is a 6" ring like the photo. I may move back to using that one.

Thank you,

Like that?

Anything that can break up the air flow between you and the microphone.

This is an example of what can happen when you don’t.

Chase was perfectly OK with that and he did many podcasts after this one.


Raw sample uploaded.
ACX check on raw file

Thinking about what you have said above, I should probably worry less about initial recording volume and a quieter recording level should require less noise reduction and just a slight bit of amplify up to specs. I can always get my chapters to pass but don’t like what it is doing to my voice.

***Whoa, just checked and the noise reduction is set to 20. You mentioned 12 being where things start to get funny.


I expect that to sound exactly like you. It does pass ACX. Anything you do from here is going to involve voice quality changes such as DeEsser or equalizer.

I did it an odd way.

I applied a home-build plugin called LF-Rolloff. It’s difficult to hear tones lower pitch than 100Hz (rumble, thunder), so LF-Rolloff deletes them all. Your microphone system creates some rumble and sub audible junk, so this has the advantage of making it easier to hit ACX Noise with few if any evil side-effects.

Then I applied another new tool I call SetRMS. It analyzes your work and forces it to the correct RMS (Loudness). This tool development is weeks old.

You got lucky because your work passes ACX just like that—barely.

I did apply Noise Reduction of the Beast, 6, 6, 6 using the silent portion at the beginning as the Profile. Nobody can hear that working except the noise magically goes down by half.

And…we’re done.

I’ll post details of these tools. You have to custom install them.

Screen Shot 2016-12-15 at 18.53.18.png

Everybody gets Noise Reduction wrong. The object is not to create Dead Blackness of Space silence between your words. Assuming a normal, gentle rain-in-the-trees hiss in the background, it’s to make it so you can’t hear it at normal listening volume.

Set my response clip so your voice is normal listening volume. Don’t change anything and listen to the silent portion at the beginning letting it play into the voice. The hiss noises should be very nearly completely gone. It’s important that it not completely vanish for another reason. ACX has a failure called Overprocessing. Odd sounding voice and dead zero noise can be cause for rejection.


Good voice if I didn’t mention that.


– You can get LF-Rolloff from me.

Decompress the ZIP archive into LF_rolloff_for_speech.xml. It’s pretty tiny.

Adding Audacity Equalization Curves
– Select something on the timeline.
– Effect > Equalization > Save/Manage Curves > Import
– Select LF_rolloff_for_speech.xml > OK. (it won’t open the ZIP. You have to decompress it)
– LF rolloff for speech now appears in the equalization preset curve list.

The Equalizer Filter Length setting should be around 5000.

– You can get SetRMS from me.

Download it and decompress it into a plain text file. That’s Nyquist programming language.

Open it in a text editor, select it all and Copy.

Select some sound > Effect > Nyquist Prompt. Paste the text in the blank Nyquist Window > OK. Mine sticks in that window time after time without pasting it again. Don’t lose any of those downloads.
Screen Shot 2016-12-15 at 19.56.38.png

The thumb and pinky gesture is the Shaka Hawaiian sign for Hang Loose. And yes, that is about the right spacing.

Screen Shot 2016-12-15 at 20.57.32.png

Wow Koz, that sounds so warm and natural compared to my edits, even on my headset at work. Thank you so much for taking the time and effort to work on this! I am going to read through your posts and cannot wait to get cracking when I get home from work. I have a much better feeling about production than yesterday, thanks for that!

I’d be tempted to apply Paul-L’s De-clicker

Paul-L's DeClicker settings used.png

Guilty as charged here. My experience with the noise reduction plug-in was just to use it as-is and boom, noise is gone. With 20 as the default number, I can see why it is so effective (but at a cost). I read some forum posts and saw that some people ran noise reduction too far and that was flagged by ACX as being overly treated. Your advice and experience is a great education and I plan to remaster all of my completed chapters.

Thanks again,

Yup, I use that one. I am trying to train myself away from it but I click. Some days are better than others but I always run it. Thanks!

I agree. I favor microphones which don’t put that grit in the sound in the first place.


Wow Koz, your treatments in this post are working well for me. I sound like me again! On top of the great plug-ins, A big takeaway is your advice on noise reduction. For my current project, I was blindly using the default setting on 20 and now I can pass ACX without any. For consistency with the earlier chapters, I am doing a light treatment, but only at a 2 setting. With that large level of NR, I was sounding a lot like HAL (for us old guys).

Thanks again and Happy Holidays!!!