ACX check

Hi there

First post! I have been recording an audio book over the last few weeks and am starting to panic about whether it will meet the submissions requirements for ACX. I’ve read a few posts on this forum and have applied noise reduction and used a moderate leveller. The volume sounds about right but it all seems a bit sibilant. I am worried that the quality of the original recording is not terribly good. Can anyone have a listen and let me know what they think? I am using a Yeti Blue Microphone in a home studio but keeping it well away from my laptop.

Many thanks!

We would request postings in WAV rather than MP3. MP3 creates sound damage and you’re never sure if the damage is coming from the MP3 or something the performer is doing wrong. You can post ten seconds of WAV mono on the forum. Since it’s difficult to edit an existing MP3, your master archive works should all be in WAV. You can make a WAV into anything.

Very well done, but I think you’re worried about the wrong things. You made the already OK noise floor much lower creating the “inky blackness of space” effect, but ignored the blue wave peaks which are too high.

Measure the peaks:
Select the whole clip by clicking just above MUTE.
Effect > Amplify. Read the top number, don’t apply the tool. It should read between 3.5db and 3db.

Measure the noise:
Drag-select between 2 and 3.5 seconds.
Analyze > Contrast: Measure Selection. It should read lower noise (higher number) than 60dB. Stunning low noise values are not automatically valuable. You never had anybody reading a story to you from a black vacuum. ACX is not impressed with extreme numbers. Unusual numbers usually cause other sound problems.

Measure the Loudness (RMS)
Select the whole clip by clicking just above MUTE.
Analyze > Contrast: Measure Selection. The value range is 18dB to 23dB. This one is fuzzy and changes with the number of pauses and how theatric you are.

So that’s how to measure the three numbers yourself. My opinion is you have a good enough “raw” production that it should be possible to conform to audiobook standards with very little work. Probably a good thing so you don’t have to start over…


I just went down to listen to your original noise. It’s not clothing rustling or wind in the trees. It’s computer whine noise similar to this:

That explains why it needs to go. Is this a USB microphone or USB MicPre or other similar system?


Thanks for that Koz. It’s reassuring to know I won’t have to rerecord the whole thing!

I’m not quite sure I understand about the peaks. I checked the amplitude as you suggested and it says amplification 1.0 rather than between 3.5 to 3.0db. How do I alter it to get it into the right frame?

The unedited noise levels all come up within range; an average of -19.6db. The background noise is coming in at -57.2, below the -60db threshold. Does that mean I shouldn’t apply any noise reduction at all? It sounds quite hissy to me.

Appreciate the help, by the way. It’s a weight of my mind!

Yes, it is a USB microphone, but I use a long cable and the computer is on the other side of the room, on a different surface. How would I go about getting rid of the sound? I take it I do need some kind of noise reduction then?

I’m compiling a list of patches I would make. Yes, we get a sinking tummy feeling when somebody opens a posting with “I’ve been recording for six months…” I favor patches that are so gentle/wimpy they’re almost seem not worth the effort, but your work is just not that far off. I could listen to that voice. Tell me all about those gales.

You’re very kind! And that’s the least interesting sentence in the book!

Import the unimproved clip.

Noise Reduction:
Drag-select the “silent” area between 2 and 3.5 seconds.
Effect > Noise Removal: Profile.
Select the whole clip (click just above MUTE).
Effect > Noise Removal:
— Reduction 6dB
— Sensitivity 0
— Smoothing 150
— Attack 0.15

Blue Wave Peak Control:
Select the whole clip.
Effect > Normalize:
— [X] Remove DC
— [X] Normalize to -3.2dB

It looks imposing written out like that, but it’s only two tools and set for gentle corrections. 6dB Noise Reduction almost seems not worth the effort, but I pushed that background whine down beyond audibility.

Normalize to 3.2dB seems entirely compulsive, but it turns out that solves a WAV to MP3 conversion problem that 3.0 doesn’t.

Note your voice didn’t change a bit. Try the 3 tests I wrote about in that other message.

You didn’t respond to my comment of doing all your work in WAV. It’s all in MP3, isn’t it?


The background noise is coming in at -57.2, below the -60db threshold.

That takes a bit of getting used to. Zero is maximum loudness. That’s where the digital system runs out of numbers and doesn’t get any louder. That’s why the sound meters have -0- all the way to the right.

That makes -57.2 louder than -60. It fails.


That does sound much better. The background hum is negligible and it gets rid of the inky black effect you were talking about. Thank you for that!

I had a go at chapters one and six and did your three tests.

The contrast for the background noise is 69.6 which is good. The overall volume comes in at 25.1 for the first chapter and 26.4 for the sixth. This is a little outside the 18 to 23db range. I think the volume overall for both chapters is quieter than the average for the sample I gave you. Chapter One unedited comes in at 22.5db overall and Chapter Six comes in at 24.5db. Should I adjust the figure for “normalize” to compensate?

I am editing in Wav by the way. I recorded all the original files using Audacity so it seemed sensible. I was just a little foxed by your 2MB limit.


Should I adjust the figure for “normalize” to compensate?

No. That would be too easy. You have to maintain the two ends of the performance, maximum blue wave peaks and minimum background noise but change the middle, overall loudness (RMS). You can do that with the compressor tool. This is where people start chasing their tail. Compressor throws off noise and fixing noise throws off…etc.

Import the chapter with the worst RMS reading. 25dB was it?
Effect > Compressor: 12dB, 40dB, 2:1, 0.2sec, 1 sec.
Effect Normalize: [X] Remove DC, [X] Normalize to -3.2.

Run the three tests again. Make sure that corrected chapter isn’t audibly different from the others.


I tried the settings you suggested and that got it down to 24db. I then tried upping the threshold on the compressor to 14, 16 and then 18db. That finally got the RMS to 22.9db. Now it sounds more or less the same as the edited sample.

I presume I should always set the Normalize to -3.2db and just up the threshold on the compressor as appropriate, with minor tweaks to the noise reduction. The one I’ve just looked at is probably the worst so hopefully none of the other chapters will need anything more drastic doing.

I’m just glad I don’t have to do any rerecording!

You can also get a boost by changing the ratio. 2:1 > 2.5:1, etc. But I agree. I made a correction for someone else with a threshold at -20 and that seemed to sound OK.

‘No Peaks Higher than 3.0’ is burned into the ACX specs. The 3.2 goal number is to avoid conflicts when converting from WAV (high accuracy) to MP3 (sloppy accuracy). You can’t miss it by much the other direction, either. The goal is to shoot live with peaks around -6 to avoid accidental overload which can create permanent, harsh damage, but the delivery spec is -3. So you will almost always be applying some correction. That normalize setting is one of the most benign things you can do. The Remove DC part of the setting is to guard against accidental Hardware damage. Rather than going down the crazy pathway to fix it for real, this is a convenient software patch. Should you actually have DC problems, it can make editing a nightmare.

In general, you should generate your correction suite for application to the original WAV, not continuously correct the last correction.


You shouldn’t need minor tweaks to noise reduction. That’s dangerous. Just enough compression to pass the RMS test and then test noise to make sure it passes -60. I think the noise values I picked should get you well below that. Noise Reduction correction numbers much past 12 indicate something wrong. Some posters with severe noise problems have corrections in the 20 range. Those are the people with Science Fiction voices and will, in general, never pass compliance.

The computer whine noise is coming from inside your microphone. It’s not room noise coming from the fan in the computer. It’s a serious problem of USB microphones and microphone systems with no good solution. Some microphone/computer combinations are unusable because of that.


The highest noise I got was -64db which doesn’t sound too bad.

The microphone’s a bit of a bind, but never mind. Serves me right for valuing convenience over quality!

Anyway, thanks for all your advice. It’s much appreciated. I think I have a much clearer understanding of how it all works now.

All the best


Well, I got through the ACX check first time without any queries so I hust wanted to say thank you for the help and advice.

It took a lot of effort to get all the chapters sounding the same, but I stuck with the Equalisation/Noise Reduction/Compressor/Normalize routine recommended on here and it got me through. :smiley:

I had my first stars on today and the production quality got four. Not bad, considering I recorded it all in my bedroom!

Congratulations :slight_smile:

I had my first stars on today and the production quality got four. Not bad, considering I recorded it all in my bedroom!

Can you post a screen grab or other indication of that Audible message? None of us are readers and the only way we have to know what the companies are thinking is when somebody actually experiences it and posts here.

We have a pretty good idea of how ACX works, but Audible is a mystery.


It was a customer rating rather than anything official. But it made my day anyway!

I uploaded the files and submitted them to ACX on Wed 10th June. It passed the check on Monday 15th (the status changed to HEADED TO RETAIL on the ACX site). The book was finally released this Tuesday (23rd) and that was when I got the formal email saying it had been published on Audible and ITunes. It took a further day or so for the audiobook to synchronise with the existing ebook/paperback on Amazon. The book appeared on the New Releases page on Audible UK and US but the first sales figures didn’t arrive until yesterday (Thursday). There seems to be a 48 hour delay in reporting - but I sold 9 copies on the first day, which I was more than happy with. ACX then sent me a few promotional codes, to help out with reviews etc.

Other than the fact of passing the quality control, I haven’t heard anything from ACX or Audible regarding the production itself. They do put up a sample of the book on the site for people to listen to though.

I haven’t heard anything from ACX or Audible regarding the production itself.

You probably won’t. They’re just concerned that you have a marketable product and then after that it’s between you and the rights-holder. ACX said they don’t listen to it.

ACX will get into theatrical quality control as part of standards by telling you that your voice quality sounds crushed and overprocessed.

It works like this: Somebody will submit a home recording with fan, hiss and other background noises marring the show. ACX bounces the show as too noisy (Their robot does that. Doesn’t have to be a human).

The performer always overreacts and noise-reduces the show so much that the voice quality turns to garbage. ACX bounces that as “Overprocessed.”

I could write the script.