ACX-Check

I have now become familiar with ACX-Check and use it all of the time, which enables me to always get passes from ACX upon submission.

Now it appears that ACX themselves have produced what I take to be a copy of yours or something to frustrate yours. Are you aware of it, and can I continue to use ACX-Check successfully.

For your interest I now have a proven Macro which passes 99% of the time.:-
Normalise>Equalisation>Compressor>Normalise>Limiter>Compressor>Limiter… If RMS is too low I repeat Compressor>Limiter which always works

The results get rave reviews from my authors

Graham Spiers (recording as Martin Hussingtree)

Our ACX plug-in is a loose re-implementation of their “robot assessor”. Their automated loudness measurement is more sophisticated than our ACX Check plug-in, in that it automatically removes long silences prior to measuring. It also checks that the loudness is reasonably consistent throughout the file, whereas our plug-in just takes an average measurement. Their “robot” was not publicly available.

Are you saying that they now provide a measurement tool that can be downloaded? If so, please post a link.

Steve

Here is a copy and paste of their recent announcement:-

ACX Audio Analysis Tool FAQ’s

Q: What is the Audio Analysis tool?
A:The Audio Analysis tool is a tool that offers ACX users immediate feedback and real-time insights on how certain elements of their uploaded audio such as RMS, peak levels, or sample rate, measure up to ACX’s Submission Requirements, right in Production Manager—without having to wait for QA review.

Q: Why does my Digital Audio Workstation (DAW) show different file measurements than the Audio Analysis tool?
A: RMS values can be measured using multiple methods and there can be many variations between soundwaves. As a result, RMS or Peak readings may differ depending on your DAW, the plugin/meter used, and the reference wave used by the manufacturer when calibrating the statistics plugin.

Q: How does the Audio Analysis tool work?
A: When you upload audio files for your ACX project—starting with the 15-minute checkpoint—the Audio Analysis tool will run automatic checks for RMS, peak levels, bitrate, bitrate method, sample rate, mixed channels, and duplicate files. The tool flags the affected files and identifies the specific issues, and users have the option of downloading a report with more detailed information.

Q: Can an author see the results of the Audio Analysis tool?
A: No. Audio Analysis tool results will be visible to producers and DIY Rights Holders only—authors in contract with an ACX producer will not be shown these results.

Q: Will I be penalized for submitting audio files that contain errors identified by the Audio Analysis tool?
A: The Audio Analysis tool provides warnings that the uploaded audio files would not meet our submission guidelines if produced at this standard. In this initial phase of the tool’s rollout, the feature does not block production of audiobooks if it indicates an issue, however, producers should be prepared to receive feedback from QA on the indicated issues and will be required to make the necessary corrections before the audiobook can be published.

Q: How can I fix the errors identified by the Audio Analysis tool?
A: The report provided by the Audio Analysis tool includes descriptions of the issues to be fixed, along with links to ACX Submission Guidelines and ACX Reference Guides and other educational resources that can assist you in correcting the issues outlined in the report. Additional educational resources for recording, editing, mastering, encoding, and asset delivery can be found in articles at blog.acx.com and in video form at acxuniversity.com.

Q: Can I dispute the errors identified by the Audio Analysis tool?
A: No. While we are constantly working to evolve our technology to hone the capabilities of the Audio Analysis and other tools, we are confident in its accuracy and ability to detect common audio errors.

Q: Where can I find the manuscript for this audiobook?
A: Once the Rights Holder has uploaded the manuscript to the project, please follow the steps below to access the manuscript:

While logged into your ACX account, click Projects in the upper left corner - then, click In Production.
Locate and click the name of your book in blue.
Select the Production Manager tab.
Navigate to the Production Resources tab.
Click Download Manuscript in blue.

Q: Where can I find the Opening and Closing Credits for the audiobook?
A: With the addition of the Audio Analysis tool, the Production Manager contains more information than ever. In order to streamline the interface, we have introduced the Production Resources tab, which will contain the Opening and Closing credits text, along with the project manuscript and the audio for the 15-minute checkpoint.

Q: Why was my audiobook rejected if Audio Analysis did not show any errors?
A: The Audio Analysis tool is an automated system and cannot detect all potential errors in an audio file. Certain more subjective issues, such as editing, extraneous noise, dynamics, and spacing cannot yet be detected by the Audio Analysis tool and will be caught instead during review by our QA team after submission.

My question remains…Will ACX-Check effectively do the same thing? Do I need to change anything?

Thanks for your continuing assistance

Graham Spiers

Where is the “ACX Audio Analysis tool”? I’ve not seen it, so I only know as much as you have written about what it is and what it does.

Below I have reproduced the article that announced the new regime:-

ACX requires the narrator to submit a 15 minute sample, in contractual terms, to commit the author to accepting the standard of work offered by the narrator. The legal principle of Offer & Acceptance = Contract applies in all law, based on the ancient Common Law, which all English speaking nations subscribe to. The idea being that ACX cannot be responsible in any dispute.

As I have succesfully used, relied upon ACX-Check in narrating and recording novels, so far, I assume that their new tool will not be any more demanding. But if they throw up some extra-over provision, then I will come back to you for further advice. I don’t expect to be submitting anything for them to try this out until Mid-Late March. Whatever the outcome I will let you know.

Feedback Without Distortion: Audio Analysis is Here
Posted on February 20, 2020 by Emily Curran | Leave a comment

Today, the ACX team is excited to announce the launch of a new feature available in Production Manager: it can analyze all your audio files, let you know if they meet ACX Standards, and give you a precise report on the changes that need to be made, all before you ever submit your project to QA—it’s the Audio Analysis Tool!

This feature will be accessible to producers and DIY authors on all new ACX projects. Now when you upload audio to ACX—starting with the 15-minute checkpoint—you’ll get an immediate report on seven common audio issues:
RMS Sample Rate
Peak Levels Mixed Channels
Bitrate Duplicate Files
Bitrate Method

That means no more waiting for the book to go through QA to learn you have one file in stereo and having to resubmit the whole project over again. Now you can find out right away and quickly make the required adjustments. The report contains precise indications on what needs to be adjusted and by how much, with links to helpful resources on how to do it. Don’t worry—all ACX audiobooks will get a final listen from our QA team before going on sale, and they’ll be on the lookout for spacing, noise floor, and other issues that can’t be detected by the tool—only now the process will be able to move a little easier, with smoother production timelines for all.

To give you some time to get used to this new feature, we’re accepting all audiobook submissions, even if Audio Analysis identifies errors within your files. When the feature launches fully, any issues detected by Audio Analysis will need to be corrected before the project can be submitted for Quality Assurance. This initial phase will give you time to identify recurring issues in your productions and make the necessary adjustments to your workflow without impacting your ability to submit audio and receive QA feedback.

We’re excited for this feature and we hope you are, too—we hope the earlier feedback and additional insights will help you improve you skills and setup, and result in a smoother production experience for all. And as always, if you have any questions about the feature, the answers can be found in our Help Center.

This looks like the source of the article … Feedback Without Distortion: Audio Analysis is Here | ACX

Their new tool is not available to the public … https://audible-acx.custhelp.com/app/answers/detail/a_id/10020/kw/audio%20analysis

Yes, if you read the posts above, they are copied out for you.

You have not added anything to the situation, but thanks for trying

I’ve added a concise summary & links to the source material.

it appears that ACX themselves have produced what I take to be a copy of yours or something to frustrate yours.

Historically, ACX doesn’t even acknowledge our existence. Our only direct contact with them didn’t go well. We spent most of the time talking past each other.

I suspect strongly they did it the same reason we did. There were no rapid, convenient tools to check your show. I produced a useful collection of “regular” Audacity analysis tools which was time consuming, awkward and inconvenient (I think I finally deleted the list—it was just too painful).

They did it because their evaluation method always involved an expensive human and everybody with respiration and a pulse is trying to read for audiobooks. I can imagine the evaluators dreading to come to work.

They increased the climb to post. The old requirement was one to five minutes. Even I could make that work. Fifteen without screwing up is a lot harder and a lot easier to analyze for acceptance. Their submission system has always been post-only. No stand-alone or remote access tools.

I’m going to go read that again.

Koz

Here it is. Buried in the justifiably dead archives. This is what ACX Check used to look like. Apparently there was an early ACX-Check app and it wasn’t entirely stable.


ACX Audiobook Testing

You can do ACX AudioBook acceptance testing with normal Audacity tools. This will only test against technical standards, not theatrical.

Manual tools can be handy if the automated Audacity conformance tool ACX Check fails or gives irrational answers.

Notes

— You should do these tests on a simple 10 second sound clip before you read long works.

— The bracket values are ACX AudioBook standards.

— Analyze > Contrast only works with a mono sound track. Contrast retains the last value, so be sure to press “Measure Selection” each time for a fresh reading.

— Effect > Amplify will not give a negative number. That’s OK. The number itself is valid.

_________________________________

— Measure Noise [-60dB or lower]

Drag-Select room tone or a pure, quiet background portion of your clip.

Analyze > Contrast > Foreground > Measure Selection (Read the number) > Close

— Measure RMS (loudness). [Between -18dB and -23dB]

Select the whole clip by clicking just above MUTE.

Analyze > Contrast > Foreground > Measure Selection (Read the number) > Close

— Measure Peak Sound. [-3.0dB or lower]
Select the whole clip by clicking just above MUTE.

Effect > Amplify > Read the number and CANCEL. Do Not Apply the Tool!!

Edit > UNDO if you apply it by accident and the blue waves change size.

— Help

You can ask questions on the Audacity Help Forum.

The forum is moderated, so even if you log in, your question may not appear immediately. Don’t double post.

2015-05-23
2016-06-14

Hi Koz

That was a fantastically interesting and most helpful reply. Very grateful.

Graham Spiers (recording as Martin Hussingtree)

For your interest I now have a proven Macro which passes 99% of the time.:-
Normalise>Equalisation>Compressor>Normalise>Limiter>Compressor>Limiter… If RMS is too low I repeat Compressor>Limiter which always works

Where is the Macro or what are the tool settings? There is no Equalization any more, so this is out of date?

If the extra Compressor>Limiter pushes you from 99% to 100%, why not include it in the Macro?

Koz

Hi Koz

I am still using Audacity 2.3.0. From what I read on here changing to 2.3.3 at this time does not help me. Equalization as it is works fine. If it ain’t broke, don’t fix it. My understnading is that the “old” Equalization is one of the two options.

Making a Macro is something that I have not done yet, although the tried and tested sequence above works very well. I am not a techy so the thought of doing so does not excite me as much as it should. Doing things manually works.If there is an idiots guide to doing a Macro kindly tell me where it is in non-techy language.

Your neanderthal friend

Graham

changing to 2.3.3 at this time does not help me.

It may not. In My Opinion you should not upgrade until you can point to a specific reason.

Equalization has been a thorn. In older versions, Low Rolloff Equalization came with variations, sliders and settings. You had to pay attention. The new rumble filter (Low Rolloff) is named the same, but it runs in the newer “Filter Curve” tool and has no options or variables. So far so good. It also comes with a name nobody can find and a much stickier problem that under Macro control, it doesn’t always call the correct curve. Not something you’d want to be running hidden in the background. I understand there’s a new version in the works.

So you’re running all those tools manually…

We publish a grandly named Audiobook Suite of tools. “Suite” because you shouldn’t run them out of order, add any, or leave any out.

Mastering.png
That’s a cutting from the much longer detailed instructions.


If you recorded in a nice quiet environment, that will produce an audiobook specification sound file. It will hit Peak and RMS right in the middle every time. The only variation is if ACX changes the specifications slightly.

The main difference between the older tools and ACX-Check and Audiobook Mastering is an acknowledgement of what “RMS (loudness)” even is. Your collection is what happens when you do it brute force without RMS tools. That long thing I published up the thread is what happens to ACX Check without convenient RMS tools.

Both older services work. That’s another fuzzy rule. If you have something that works, hold onto it with white knuckles, and DO NOT change anything in the middle of a book.

I don’t know that there is a “Macros for Dummies.” It’s simplified, but you are managing a programming language with all the nutty things that can happen when you do.

I have to drop for a while.

https://manual.audacityteam.org/man/macros.html

Koz

Normalise>Equalisation>Compressor>Normalise>Limiter>Compressor>Limiter.

Do all those duplications operate with the same settings? For example, do both Compressors operate with the same sliders and checks? I ask that because straight, simple compressor and limiter are downward-only tools. They don’t boost anything and adding another pass of regular Compressor and Limiter to the chain won’t help—depending on what failed. Home readers are always too quiet.

The results get rave reviews from my authors

What do you do with new authors who don’t pass noise? Nobody reading from home passes noise. Homes quiet enough to pass are very rare.

This from counting the number of forum new readers with exactly the same complaint. “My readings are too quiet, I can’t pass all three ACX values at the same time, and my submissions are too noisy.”

Koz

I have posted files since they implemented this automated check. It found no errors in any of my files.

I used the suite of Rolloff, RMSNormalize, Limiter, and then ACXCheck and as long as the files passed and on listening, I found them acceptable, I uploaded them. I am still waiting on the manual QA process, so I will let you know more.

I have posted files since they implemented this automated check.

Tests or the real thing? A whole book?

I read that three times. You posted chapters to ACX and their new robot quality control didn’t find anything wrong. They always had automated acceptance at the beginning, but this new one is different and we assume “better.” We know from postings that the ACX one takes into account the silences between words and theatrical pauses. Audacity Audiobook Mastering doesn’t, but we still expect to get numbers close to each other. It’s reassuring that this seems to be the case.

I am still waiting on the manual QA process

You’re waiting for ACX’s Human Quality Control. Yes. This is the step you spend on the edge of your seat waiting for them to object to some announcing oddity or sound damage only obvious to a human.

If they do accept you, you will get a congratulations message and other brief comments about what’s going to happen next. Please post that if it arrives. If they reject you, they usually list the parts they didn’t like, but traditionally, they won’t engage with you after that. As I’ve posted before, it’s not their job to troubleshoot your studio.

It does occasionally cause confusion if you have multiple problems because they may not specifically complain about all of them. Only the worst ones. Also, again Noise pops up, you two may get conflicting analysis if your noise passes the -60dB limit by multiple different measures, but they bounce you anyway. Some noises such as the USB microphone “frying mosquitoes” sound is still objectionable even in the background. That’s the “baby screaming on a jet” effect. It’s not very loud, but just try to sleep through it. There was a recent “Sandra and Woo” cartoon where one of the characters got stuck on a plane bound for a colicky baby convention. He offered to deplane before they landed.

ACX’s overarching goal is to reduce distractions to zero. I believe that’s still their goal.

I see ACX’s postings are a good deal more interested in establishing the relationships between authors, performers and ACX. One can only imagine the problems if they published something where the author had no idea what was happening.

Koz

The robot seems to check each file during upload and the Audio Analysis tab is present and has feedback after the first upload.
I didn’t get any hits so I do not know if it is a cumulative total.
AudioAnalysis.png
It always had that same message from the start to the finish.

So that’s what success looks like now.

What was the delay submission to finish?

The robot seems to check each file during upload

I’m not shocked. My impression is of a hierarchical testing structure. If you fail basic file or noise limits, you go to the office where people give you a brief explanation of what you did wrong and point you to prebaked solutions. They won’t entertain conflicts or try to fix your studio with you.

If your files are OK, only then do the serious Quality Assurance people listen to it for theatrical errors. Are you recording in a bare room with echoes, do you stutter or split when you talk? I see by their postings that they expect to lift humans out of these jobs as well. Algorithmic analysis of human speech is not that hard any more.

And just to inject a ray of sunshine into this process, Text to Speech software is getting better and better. That’s the end of the QA office.

Koz