ACX say my RMS is too low

I am on Windows 10 and the latest version of Audacity.

Until they changed ACX requirements to LAME Encoder, I was happily satisfying their specification of between -23dB and -18dB by using ACX Check. They accepted 2 long audiobooks on that basis.

Their guidance on the change was to use another piece of software which was not Audacity. The advice on here from Audacity commentators was that the Lame Encoder and ACX Check would do the same job. No need to use WAV and the other software. There is no change in the ACX specification.

So I continued. They rejected my first effort. On the 22 Chapters submitted the RMS was between -18.9 and -22.9. They suggested Normalising to -3.0 aiming for about -22.#.

My latest attempt had a range of -21.3 to -22.9. Peak level using the Limiter was consistently -3.7 and -3.8.

ACX say that the RMS is too low and needs to be louder. They do not say what numbers to aim for which would be far more helpful. Could one of you kindly tell me what numbers might satisfy them.

Grayspires (Martin Hussingtree)

[u]AXC requirements[/u]

So I continued. They rejected my first effort. On the 22 Chapters submitted the RMS was between -18.9 and -22.9. They suggested Normalising to -3.0 aiming for about -22.#.

Strange… That should be OK.

My latest attempt had a range of -21.3 to -22.9. Peak level using the Limiter was consistently -3.7 and -3.8.

Your peaks are a little low. They say -3dB but they don’t give a range or tolerance. If you amplify by 0.7 or 0.8dB, both the peak and RMS will be boosted by the same amount. (But less than 1dB is not THAT much difference.)

You can bring-up the RMS level by amplifying again, then run the limiter again to bring the peaks back down. Or, you can use the Envelope tool to fade-up the quieter parts. In fact, that may be the problem… Maybe your overall RMS level is OK, but there are some too-quiet parts???

Until they changed ACX requirements to LAME Encoder, I was happily satisfying their specification of between -23dB and -18dB by using ACX Check. They accepted 2 long audiobooks on that basis.

When you convert to MP3 the RMS shouldn’t change, at least not significantly. However the wave shape does change. Some peaks get higher and some get lower. The new highest-peaks can be 1dB or so higher, especially if you’ve used limiting to “artificially” limit the peaks.

So, you might have to go-back and re-adjust the levels and make a new MP3. But DON’T edit the MP3! Go back to the original uncompressed file. When you open an MP3 in Audacity (or any normal audio editor) it gets decompressed. If you re-export as MP3 you are going through another generation of lossy compression.

Until they changed ACX requirements to LAME Encoder

What does that mean? Who changed what?

Their guidance on the change was to use another piece of software which was not Audacity.

Such as what? We’re not delicate. We’ve been known to recommend other software when appropriate.

-22.9

That’s dangerous. 0.1dB is up for interpretation and normal error failures. Just simple conversion to MP3 can cause errors that big.

Whose mastering are you using? We published a simple mastering technique that always puts your passages directly in the middle of RMS specifications. It appears to me you’re trying to read directly into ACX. You can do that, but in my opinion, that can lead to “brittle” readings, where if you do anything wrong at all, you fail.

It gets worse. ACX Check is our approximation of what ACX is actually doing based on their published specification, performance successes, and failures. It’s been just fine for a majority of users, but you can cause errors.

All that and ACX works in two levels. The Robot checks the same thing that ACX Check does. But then you have to pass Human Quality Control and it seems that may be where you’re running into troubles.

Congrats on your first two books.

I know this is out of order, but can you post a 20 second sound test?

http://www.kozco.com/tech/audacity/TestClip/Record_A_Clip.html

Don’t change anything. Read, export and post.

Please comment on the questions in the above.

Thanks,

Koz

Thanks for your kindness.

Please find attached the aup file you asked for. I hope I have done what you asked. Nerve wracking


Best wishes


Graham
AUDACITY TEST.aup (1.55 KB)

Good idea and concept, messy on execution. The idea is to record the work (and cut it down to size if needed—that much editing is allowed). Then File > Export > Export as WAV.


Screen Shot 2019-03-09 at 12.29.05.png
Audacity does not Save sound files. You sent an Audacity Project Manager Text File (AUP) by accident.

The illustration is what it looks like on a Mac, your Export will look different.

Nobody will stop breathing if you get lost here and there. Just post where you are and a forum elf will help you.

Koz

Hi Koz

Sorry for sending you the wrong file. Please find attached the WAV exported untouched file.


I have asked ACX to tell me what they want

Thank you for posting your sound sample.

I have asked ACX to tell me what they want

They do have postings with reading requirements.

https://www.acx.com/help/acx-audio-submission-requirements/201456300

But I don’t think that’s the problem. I was able to apply our ACX Mastering tools and I got your submission to easily meet ACX technical guidelines. But. It sounds like you’re reading into a milk jug. Did you get a new computer between your first two books and now?

Describe your microphone and recording system. Include part numbers where possible so we can look everything up on-line.

If you’re using a computer for recording, it may be trying to help you by automatically processing your sound in the background, similar to how cellphones work. This is one reason we wanted you to post a “clean” sound sample, so we can tell if there’s sound damage from somewhere else.

Koz

Do you use Skype or other chat or conference application?

Games?

Koz

Thank you for your continuing interest and tolerance.

I am saddened by your unkind reference to my large purpose built acoustic foam recording booth as a milk jug. I have been trying to send you PDF photos but I cannot get your website to accept them as an attachment. I have a friend who is a BBC sound engineer who has been most complimentary.

My condenser microphone is a Prosound attached to a Behringer U-Phoria with a pop screen. My “studio” has soft walls and floor. I live overlooking open countryside so I can even shut out birds chirping. I get no Noise floor problems. That has all been constant. My computer is a Surface Pro running Windows 10.

Since ACX accepting my first 2 long novels, the only changes have been their introduction of the need for a LAME Encoder and 192 bit constant etc… Their advice which I have not taken up is to export as WAV to an alternative software to Audacity. I am happy with Audacity and ACX Check. I cannot understand why their suggestion excluded Audacity given its universality. Maybe you should speak to them.

Following my questioning their advice on here, as you will see from my earlier post, you or your colleagues saw no reason to change from Audacity. So I didn’t.

Then recently you asked me to send you a raw WAV file which I did and you likened it to a milk jug. I have no previous experience of WAV. You would probably regard me as electronically Neanderthal.

I have not yet had a reply from ACX asking them specifically what they need me to do, as I have complied with their specification, and which you have now confirmed that I have.

On my equipment everything sounds OK, even that of my sound engineer friend, and he is high end fussy.

I am sure we could communicate with WhatsApp or Skype if you wanted, but I have never done either. I am 74 years old and not as mentally nimble as I would like.

I have wondered about your RMS Normalise thing, but until ACX tell me what they want the RMS to be there seems little point.

I hope I am not being too much of a nuisance

Kindest best wishes

Graham

I am saddened by your unkind reference to my large purpose built acoustic foam recording booth as a milk jug.

It’s not the environment. That seems fine. It sounds like electronic processing. This is typical of using a computer whose Echo Cancellation or Auto Noise Removal process is running.

When you intake breath before your first word, it doesn’t sound like a light, airy, healthy breath. It sounds like breathing into a wine glass or milk jug. That is a signature of voice processing.

Since you’re on Windows, have a look at this posting.

https://manual.audacityteam.org/man/faq_recording_troubleshooting.html#enhancements

As far as the rest of the complaint, we posted an Audiobook Mastering process that guarantees both RMS and Peak conformance with minimal sound quality change.

With the quiet environment afforded by your sound booth, a sound clip should sail right through.

Koz

A note on Noise. ACX Check requires about a 3/4 second of clean background sound (Room Tone) to give accurate readings. It’s unlikely to find that in normal conversational reading and will give an artificially high value.

ACX requires well-behaved Room Tone at the beginning and end of each chapter, so if ACX Check fails, so will the submission—oddly, not because of poor sound quality, but a bookkeeping error. Not enough Room Tone.


Screen Shot 2019-03-12 at 09.37.02.png
Koz

And with regard to your test clip. I applied the Mastering tools. First three readings and sentence 2/3 down.


Screen Shot 2019-03-12 at 11.37.41.png

Note Peak is quieter than the required -3dB and RMS in in the middle of -18dB to -23dB. Noise passes, too at -61dB with a 60dB limit, but 1dB is not much and I think we’re going to have troubles with that when we resolve your voice quality problem.

Koz

They say -3dB but they don’t give a range or tolerance.

ACX Technical Specifications have two limits and one range. Peak can be no louder than -3dB and Noise can be no louder than -60dB.

RMS or loudness is a range. -18dB to -23dB. Many natural, theatrical readings exceed -3dB on peaks when they are adjusted for the middle of the RMS range. That’s why the ACX Mastering we published sets loudness first with RMS Normalize and then gently pushes the peaks into compliance with Limiter.

Mastering guarantees Peak and RMS leaving Noise. If you shoot in a quiet, echo-free environment, you can pass all three.

Koz

Sorry to do this to you, but it is getting tedious.

All I want to know is what settings to use to give ACX what they want. Hopefully I won’t have to re-record 22 Chapters.


I have now received a belated, seemingly annoyed, reply from ACX, swamping me with data which makes my head spin. I have copied and pasted that response here in the hope that you might kindly give me a simple layman’s answer as to what I should do.


There are pages of it. It seems to me that they have an automatic WavStat machine which analyses each Chapter every 30 seconds which is why there is so much of it. I may be wrong, but you might recognise what I am seeing. Rather than send you all of his data, I have copied as best I can information numbers on another computer what I take to be a summary, hoping that that will mean something to you.


Seriously, I really do appreciate your interest, tolerance and assistance.


Best wishes


Graham




‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’‘’’

Taddeo, Michael
Attachments
Wed, Mar 13, 6:52 PM (19 hours ago)
to me,

Hi Graham,

Attached here are the audio analysis statistics for the currently uploaded files for “More Dark Drink and Conversation (In Mulligans Book 2)”.

ACX requires that all submitted files measure at an RMS level between -23dB and -18dB, with a maximum peak level of -3dB.

As an example, the file: “NEW CHAPTER FIFTEEN LAME ACX MCCD.mp3” measures at -35dB RMS, and -19dB Peak.

To remedy this, I recommend to ‘normalize’ this file to -3dB (peak). This will increase the level of the entire audio file to the provided level (-3dB Peak).

Please note “ACX Check” is a third-party plug-in not developed by ACX.

We recommend using the “Wave Stats” plugin, which can provide reliable peak and RMS data for up to 30 second windows of a file. After installation, it can be found here:

Analyze>WaveStats

As an additional reference, we’ve put together a sample pack that contains samples of both well and poorly produced audio, as well as ideal examples from each stage in the production process. The included files are WAV files (not encoded as mp3) that you can load into your software to see how they look and sound, how high they peak on your meters, etc.

utomatically download the zip file by clicking here.

Feel free to reach out with any other questions.

Kind regards,

Michael Taddeo

Asset Specialist & Production Advisor, ACX

audible, an amazon company

1 Washington Park, Newark, NJ 07102


FREQUENCY
(kHz)

BIT RATE
(kbps)

AVG RMS (dB)

Pk lev dB

44.1

192

-31

-15.37

44.1

192

-27.86

-11.27

44.1

192

-23.37

-8.04

44.1

192

-26.86

-12.04

44.1

192

-22.53

-7.04

44.1

192

-20.09

-4.04

44.1

192

-30.33

-13.37

44.1

192

-30.17

-14.95

44.1

192

-17.96

-3.4

44.1

192

-35

-19.04

44.1

192

-20.06

-3.99

44.1

192

-20.01

-4.04

44.1

192

-20.1

-4.04

44.1

192

-34.45

-18.96

44.1

192

-30.67

-15.06

44.1

192

-35.43

-19.47

44.1

192

-35.86

-19.86

44.1

192

-28.17

-13.04

44.1

192

-23.95

-8.03

44.1

192

-28.96

-13.04

44.1

192

-32.4

-17.03

44.1

192

-31.36

-16.04

44.1

192

-29.37

-13.5

44.1

192

-20

-3.3

The problem with the [u]Wave Stats plug-in[/u] is that it can only analyze 30 seconds. I wonder if they are analyzing the first 30 seconds of your files chapters?

Would that explain this?

As an example, the file: “NEW CHAPTER FIFTEEN LAME ACX MCCD.mp3” measures at -35dB RMS, and -19dB Peak.

P.S.
Wave Stats is also not developed by ACX. And, I wonder if that’s what ACX is actually using…

Please note “ACX Check” is a third-party plug-in not developed by ACX.

He’s right. We designed it and It’s a reflection of their own published standards and our experiences. For the most part, if you can pass that, you can pass actual ACX acceptance.

It’s fascinating their method of testing and their published specifications don’t exactly match.

RMS (loudness) is an average. They do not publish the time over which they measure it. Ours is open-ended and theirs apparently isn’t. Good to know. I can understand that if you read expressively, you could pass one and not the other.

And thank you so much for posting the actual dialog with ACX. Maybe you don’t completely understand the instructions, but we do and it’s very helpful.

As we go…

Koz

  1. Launch Audacity

  2. Import “NEW CHAPTER FIFTEEN LAME ACX MCCD.mp3”

  3. Ensure that the entire track is selected (“Ctrl + A” to Select All)

  4. “Analyze menu > Contrast”

  5. Click the first “Measure Selection” button. What does it say in the “Volume” box?

  6. Close the “Contrast” window

  7. “Effect menu > Amplify”. What does it say in the “Amplification (dB)” box?

  8. “Cancel” the Amplify effect.

Steve

I cannot get to what you want once LAME Encoder has been applied and the file EXPORTED as MP3.

I have gone back a step to the aup file R MCCD CHAPTER FIFTEEN. That enables me to do as you ask. Your instructions are clear ad understood.


The first Contrast volume measurement box gives -22.60dB


The Amplification (dB) box gives 3.774


I hope I have done what you asked me.


Perhaps in due course you will tell me the significance



Kindest regards


Gray

Perhaps not.
At step 2, I said:
2) Import “NEW CHAPTER FIFTEEN LAME ACX MCCD.mp3”


To give us the information that we require, you must “Import” the exact same MP3 file that you sent to ACX.

Sorry Steve

I think I have now done what you asked.


The contrast measurement volume box says -38.31


The amplify box says 19.484


Do they sound like what you were expecting to see?


Regards

Gray