A salutory lesson

I finally finished my audio book a few days ago and uploaded it to ACX. I used the ACX check plug in to ensure that all files passed the ACX requirements. Silly me. One feature I used to reduce the noise floor to within limits was to use the noise reduction facility in Audacity. Two days after uploading, the ACX QA guy comes back to me and says that the files have been ‘gated’ too much which has clipped some sounds and introduced ringing artifacts, whatever they are. His suggestion was ‘to reduce or remove the amount of noise reduction you are using to allow for a more natural sounding recording. Please also adjust or remove the gate settings to allow for every word to remain intact.’

There are several issues here. The first is that I sent them a sample file weeks before recording the whole book which they never bothered to reply to so I stupidly assumed it was alright. Secondly the word ‘gating’ does not appear anywhere on the ACX submission requirements page and frankly I still don’t know what it means. I have no idea whether the ringing artifacts and the clipping are actually because I used the noise reduction facility to meet tier 60db floor requirements or there are two issues here.

Finally as all I did was save the files once they passed the ACX check on Audacity I have no raw files to go back and play with so have the horrible feeling that two months of work has just one up in smoke.

I have to wonder just how many amateurs actually manage to get an audio book to meet their standards. Is it a conspiracy to drive us all into the arms of the production companies???

I have no raw files to go back and play with so have the horrible feeling that two months of work has just one up in smoke.

And that, in one sentence, is probably the worst newbie mistake. We can’t take effects out of a production. Once you burn effects and filters into your work and “step on” the original files, the corrections are permanent.

ACX Check success will only get you past the ACX Robot which blindly tests for the three technical audio specifications: Overload (Peaks), RMS (Loudness) and Noise-when-you-stop-talking. It doesn’t check content or voice quality. That comes next when a Human QC gets it.

I sent them a sample file weeks before

I don’t know where to go with that. I suspect ACX AudioBook is swamped with submissions and just can’t keep up. We get little pieces of this in our direct dealings with the company. It was difficult to get them to tell us their exact technical specifications so we could develop AC Check. It’s possible they just didn’t know and were either making it up as they went along, or had an engineer develop the system who then went to work somewhere else. We still don’t have official recognition of the work. We know by trial and error that if you can pass ACX-Check, you can pass the ACX Robot, although we do warn you the tools are not exact or perfect. Passing by some margin is recommended.

We also know that you can use the Audacity tools to force any presentation to pass ACX conformance, no matter how bad it sounds. It’s one of our fuzzy problems. How do you tell?

It’s very common knowledge that you can plunk down a Blue Yeti on the kitchen table, read an audiobook, apply a few filters and be good to go. If you’ve been reading through the Audacity audiobook postings, you know it’s a little more complicated than that.

Overapplying Noise Reduction is pretty common and yes, the symptoms are ‘talking into a wineglass’ sound with ‘tinkly bells’ sprinkled in the speech. It’s a cousin to bad cellphone sound and for the same reasons. Excessive voice processing.

It’s a little surprising they called out gating if you don’t remember putting any in. Gating is a distinctive sound. Basically, that’s a tool that only passes sounds above a certain volume. Misused, it cuts off ends of words and makes speech seem robotic and staccato.

It’s difficult across nine time zones to analyze your work, particularly if you don’t have any of the original readings left. We are, almost without doubt going to agree with ACX. Post the info from that pink band at the top of these messages, post a description of your studio including detail part numbers where applicable, do a reading according to this formula and post it.



Assuming that you did not also use a “Noise Gate” effect, you probably applied Noise Reduction too aggressively.

In several places in the ACX documentation it suggests that you should not use “noise reduction”. In fact, it is OK to use a little noise reduction, but it should not be noticeable that you have done so. If the noise level in your original recording is at such a level that you need more than a tiny amount of noise reduction, then you need to fix your recording process. High or even medium levels of noise cannot be removed effectively in “post production” without causing noticeable damage. ACX have noticed the damage cause by noise removal, and that is what they are objecting to.

Your recording room must be very quiet and substantially free of echoes.
Recording levels must be carefully set (aim for a peak level of about -6 dB).
Most vocal microphones require that you are close to the mic and use a “pop shield”.
The recording equipment must be reasonably good quality.

Always keep backups - preferably several backups. Make it a habit to export each new recording as a 32-bit float WAV file and make a second copy of it onto an external drive.

Making professional quality voice recordings is not easy. We are happy to try and help.

Good replies thank you BUT both are from experts in audio production which I am not. Some years ago I started self publishing on Amazon. There were several ways to produce Kindle files all requiring separates conversion software and a good level of technical knowledge. Now all you have to do is upload the file in Word directly. With paperbacks (CreateSpace) it was even harder with gutters/margins/headers/footers/fonts/page blanks all needing to be exactly right or it got rejected. Now you upload a Word file, it says its all wrong and ‘would you like to download a corrected version’ that CreateSpace automatically generates and the jobs done.

My first audio book was picked up by an American producer and using the raw files I made here in my office at home was able to turn them into a book that is now on sale. So it IS possible to make my basic product into a quality that ACX are happy with and I will now be going back to those guys and going down that route because basically I can’t be bothered to dig any further into this morass of jargon and extremely badly explained procedures (for an amateur) Ok I’ll lose a small amount of revenue but I might still have some hair by the end of the year.

But if my producer can do this why isn’t there some software (like on Amazon and CS) that can take raw files, filter out the noise, change the levels and peaks to meet requirements. For example I did an experiment just now and used the ‘generate silence’ feature on each gap between words on a fairly noisy track and once done it passed the AXC check and I hadn’t touched the words at all. The problem is that it is far too time consuming to manually do this on a book with 45 chapters and 90,000 words. But surely it wouldn’t be rocket science to have a feature that recognises the gaps and automatically applies silence - or am I just being a grumpy old git?

Quite correct. It is very easy to do that. A “Noise Gate” can do that.
Unfortunately, speech that is processed in that way sounds weird and unsettling. Words … become … detached … from … each … other … making … it … sound … more … like … a … machine … talking … than … an … engaging … experience … of … listening … to … a … story … being … told.
What ACX want is that the story carries across to the listener without distraction. They are quite happy for there to be a little “room tone” between the words, but not a distracting amount of noise or absolute silence.

I don’t get that - the gaps between words are already there and set by the narrator. As I said I can manually add silence in those gaps and it doesn’t distort the narration, in fact I have done so in quite a few places in the book where the noise cancellation didn’t get rid of a residual noise and to my untutored ear it sounds fine.

I don’t make the rules :wink:
ACX publish their guidelines here: https://www.acx.com/help/acx-audio-submission-requirements/201456300
That page also has links to some useful information on how to meet the requirements.

Yes I’ve read those guidelines till I’m blue in the face and with the help of this forum thought I’d got to the bottom of them. Clearly I failed. However you haven’t answered my previous point about using the ‘generate silence’ feature of Audacity in an automatic manner except to say it breaks up the audio which I observed didn’t actually seem to be true.

Inserting absolute silence between words will not be acceptable for ACX. They specify in several places that they want “room tone” as the constant background ambiance and “free of extraneous sounds”. I have previously discussed the matter of “room tone” with one of the QA managers at ACX and he confirmed that absolute silence between words will not be accepted.

the gaps between words are already there and set by the narrator.

No, they’re not. They’re set by the narrator plus the noise the microphone is making plus the noise the apartment is making. If you stopped talking and the sound on the recording went to near zero, there would be no problem. But that’s pretty rare.

As an experiment, I set up a simple microphone/mixer/computer in my quiet third bedroom and was able to produce an ACX-quality clip with simple volume changes. No noise reduction, no processing, no compression. Like I posted earlier, the prevailing wisdom is to crank out an audiobook by sitting a microphone on the kitchen table (or office desk) and go. That’s probably not going to work.

Ok I’ll lose a small amount of revenue but I might still have some hair by the end of the year.

That is one of our recommendations along with simply showing up at a studio and walking out with a perfect recording.

it breaks up the audio which I observed didn’t actually seem to be true.

How are you listening to the work? “It sounds good to me” on laptop speakers may not be good enough. They hide many sins. Simple earbuds won’t do it, either. Most of those are terrible.

That ten second sound clip is still highly recommended. We can tell you by analyzing that clip what’s likely wrong with the work.

Somebody should come up with a software package that automatically handles this. Audacity is a volunteer organization. Are you volunteering to produce that software? There’s no shortage of people volunteering us to do jobs.

“It’s too much bother” can be solved with another pathway. There are organizations that will post an audiobook with no technical standards.




Koz, thanks for all that. By the way I am listening with a rather expensive pair of headphones. As for the ‘someone coming up with the software to automate the process’ I wasn’t thinking of you guys who clearly do an outstanding job. But clearly both Amazon and CreateSpace thought it worthwhile to do it for Kindles and paperbacks so why not ACX?? - they are an Amazon company after all!! Maybe they could pay you guys to do it - there’s a business opportunity if ever I heard one.

But you’re not going to post the clip.

We have a label call 'Uncooperative Poster." Would you like that?

So far this has been a multi-chapter complaint session.


As I wrote earlier, replacing the spaces between words with absolute silence is easy. This “Noise Gate” that I wrote can do that: http://wiki.audacityteam.org/wiki/Nyquist_Effect_Plug-ins#Noise_Gate
However, converting the spaces between words to absolute silence will not be accepted by ACX,

To be blunt, Amazon do not need your audiobook. If you want Amazon to sell your audiobook, then it up to you to meet their requirements. If you can’t or won’t meet their requirements, then they won’t sell your audiobook. Perhaps that’s unfair, but that’s the way it is.

Your producer has a brain, ears and experience.

Your computer only has a CPU. It’s blind and deaf and will only do what you tell it to do. I doesn’t know what noise is, or a voice. It doesn’t feel at what level noise becomes a nuisance.

And you might think there are rules for noise, but there aren’t. -60 dB noise might sound OK in one instance and totally unacceptable in another, depending on noise contents.

It’s like cooking. A great recipe will be peanuts to an experienced cook, but someone who hasn’t baked an egg before won’t be able to execute it.

The best you can do, is post a sample and follow Koz’s recipes. They’re great, once you’ve learned to bake the egg.

Sorry everyone - I’ve tried twice to upload a sample clip. Both times I’ve had this error message:

Service Unavailable
Guru Meditation:
XID: 140924128
Varnish cache server

Whatever that means. What’s worse it then locks me out of this web site for up to 4 hours so I’m not going to try again. I have my production company now hoping to reclaim my files so fingers crossed.
However I would point out in my defence that I started this thread as a lessons learnt for any newbies hence the title not really for all you experts although as you replied, like on all forums, the thread drifted.
Anyway many thanks for all the advice but even though I’ve got two engineering degrees, chartered status and am a certified aircraft engineer who has also worked in software and electronics - I’m going to admit defeat and let the experts do my books from now on and I’ll concentrate on writing and narrating.

You could always post the sample on a filesharing site like Dropbox and then send us the link.


Funnily enough I’m using Dropbox at the moment to upload my old audio files to my producer. As I don’t need the sample analysed now then I’ll leave it unless you guys want to se it for some reason.