Perhaps a bit harsh.
An alternative view: ACX are in the business of acquiring audio books for commercial publication. Customers that purchase the audio books are likely to expect the recording quality to be of a reasonably professional standard. ACX provide recommendations so that new authors have some idea of what is required in terms of “professional” recording quality.
My view is that the numbers are a guideline, both to inform new authors, and to assist separating the wheat from the chaff.
And as I posted, the guidelines are in dB and RMS which is profoundly unhelpful if me mum is trying to produce an audiobook. I know people who are terrific readers and narrators and who wouldn’t make it past any of the technical standards. Only engineers with good reading voices need apply.
I think given a quiet recording, careful application of Chris’s Compressor will easily produce an ACX compliant presentation — repeatedly — and we can produce specific instructions to reveal whether the noise level is complaint as well. And all without signing contracts.
Sadly, fixing a high noise level gets us into the weeds again, but it’s hard to beat the simplicity of the above process. Do you think that Car Talk® clip meets ACX standards? I do. But nobody will ever know until they review it. And apparently, they won’t review it without producing a finished work and signing the contracts — loop as desired.
Koz
Which clip is that? Is it this one: Manipulating Decibels - #3 by kozikowski
Yes. That’s a segment from Car Talk show 1413. It’s a studio quiet show with no attempt at volume regulation in the podcast. Tom has the nuclear laugh and Ray tends to talk in his beer. I didn’t try the noise test. This is just a pass at Chris with the compression turned up a bit.
The peaks hover around -3 and the dark portion of the sound meter flirts with the required RMS values — if you believe the meter.
It sounds reasonable and It’s repeatable.
The history of Chris and that show goes back to an effort to make the podcast sound like the broadcast on a local radio station. For that, you crank the compression up to 0.77. No other changes are required. So 0.6 is actually looser and more expressive than the show appears in Lori (my lorry) on the way to the Tesco/7-Eleven.
@Xena In any event, we may find out shortly if Xina succeeds in the posting.
If you could post a portion of the show here that would be helpful. The work is in MP3, but a fresh, short segment would be good.
I don’t remember if we warned you about saving your work in WAV. MP3 always creates sound distortion and it doesn’t usually become a problem until you try to reedit some of your work. Show archives should always be in WAV.
Koz
I heard back from ACX, but only to tell me that my request was being escalated. Koz
I see this thread has continued on in my absence! I was out hawking storm windows, you see. 
As far as signing up with ACX, I registered on their site and instead of commissioning a narrator, I just chose the option of narrating the text myself. I was then taken to a screen where I was prompted to upload cover art and audio files and then submit them for review. I haven’t made it all the way through that screen yet, since I hit a bit of a snag with cover art and since I am working my way through each file by applying Chris’s compressor and making sure that all the tracks align sufficiently and fixing boo-boos that I missed on the previous 57 listens of the tracks (well, 57 may be a slight exaggeration but that’s what it feels like!).
I’m afraid I got a bit lost on the previous discussion of noise floor and so forth. At the risk of sounding like a complete and utter idiot, should I attempt to understand that discussion or just continue on my merry way?
I appreciate everyone’s help and attention! I thought this audiobook stuff was going to be so much simpler and quicker when I first started out, haha!
Continue on your merry way. Nothing to see here.
The point of the technical discussion is that the ACX specifications are too technical for the average reader/performer. It’s possible that a relatively simple application of Chris’s Compressor is all you need.
There’s nothing simple about what Chris’s software is doing to the show. Writing software that gracefully changes sound levels is not trivial.
I may set up to record a simple voice performance later. It would be helpful if you have a sample of work that doesn’t give away any plot lines or cliff-hangers. You can only post very brief samples on the forum, but if you contact me via email, that system can handle up to 25MB of work.
Koz
My guess is that sample will (just) fail the ACX test because it does not (quite) meet their requirements (http://www.acx.com/help/acx-audio-submission-requirements/200485520).
Each uploaded file must:
… [snip] …
Have peak values no higher than -3dB.
… [snip] …
Have a noise floor no higher than -60dB RMS.
By my measurements, the peak level is -1.9 dB which is just a little too high. That is easily fixed by limiting to -3 dB.
By my measurement, the noise floor is about -47 dB rms, which is just a little too high. This is easily fixed by applying a low frequency high-pass filter (rumble filter).
If it gets past their automated checker then they may still pass it, but I’d guess that if limited and filtered to remove rumble then it would sail through.
I’d be very interested to hear an official response from ACX.
By my measurements, the peak level is -1.0 dB which is just a little too high. That is easily fixed by limiting to -3 dB.
That happened when I bumped up the compression. The last Chris setting down is Maximum Amplitude.
By my measurement, the noise floor is about -47 dB rms, which is just a little too high.
How did you measure that?
This is easily fixed by applying a low frequency high-pass filter (rumble filter).
That’s valid. Nobody I know records voices under 100Hz. Effect > High Pass Filter: 100Hz, 24dB/Octave? 12dB/Octave?
Koz
Steve probably has fancy plug-ins 
but the methods in Audacity as shipped (as I said) are Analyze > Contrast and Analyze > Sample Data Export.
Have a noise floor no higher than -60dB RMS.
How much difference is weighting/unweighted going to make to that figure?
Is it really a “specification” without specifying weight?
Gale
Typo: should have been -1.9 dB (now corrected in my original post.
Select all, then “Effect > Amplify” shows the peak amplitude to be -1.9 dB.
or install this “show peak amplitude” plugin Peak Amplitude (Analyze plug-in) which shows the peak amplitude to be -1.87 dB.
For 100 ms window rms:
Select a “noise only” section.
Nyquist Prompt effect.
(linear-to-db (peak (rms s) ny:all))
Repeat for a couple of other “noise only” sections to check.
For a window size equal to the selection length, same procedure as above but replace the code with:
(let ((step (truncate len))
(s-sq (mult s s)))
(linear-to-db
(snd-fetch
(snd-sqrt
(snd-avg s-sq step step op-average)))))
Or as Gale suggested: “Analyze > Contrast”
Alternatively, use this “wave stats” plug-in: "Wave Stats" plug-in which gives you peak, rms, and DC offset measurements.
Interesting to note that each of these rms measurements give slightly different values, but they are all in the same ballpark.
I used 100 Hz 12 dB/Octave.
Is it really a “specification” without specifying weight?
Without a weight specification, I assume either C (minimal) or Z (no weighing).
Although it could be argued that legal and hazard specifications are always A, and those readings are always lower. In A, you might not need the High Pass Filter to squeak through. A is 20dB down at 100Hz.
Anybody write an A filter?
That’s flirting with the weeds again, particularly as we don’t (yet) have a formal noise meter.
And as I am want to continually remind people, it’s not one number. A massive thermo-chemical explosion at the beginning of a 60 minute production of dead silence will, taken as a whole, pass the ACX noise test. RMS, by definition, has a time value.
Koz
When I spoke (e-mail) to ACX a while back I was told “unweighted rms”.
Weighting would not normally make a huge difference, but DC offset will be ignored by a-weighted measurements, so that could cause a significant difference. To be a proper “specification” it should say what weighting and what window size.
C weighing doesn’t go all the way down, either. Odd they picked Z. Maybe they were trying to filter out (to to speak) all those people with significant technical problems. If I was going to choose a test to trip up 90% of voice producers, that would be it. Can your office off the hallway just past the loo pass that noise test? Nobody can pass -60dB, Z-Weighing.
That and “You Failed” doesn’t tell you where to go after that, even if you know what failed. “Your RMS is too high!”
…ok…?? [said with a rising inflection]
Apply noise removal and post on the Audacity Forum when that doesn’t work.
Koz
Weighting would not normally make a huge difference
Not on voices, no, but dance clubs live on the difference. All that shirt-moving bass assault doesn’t register.
Koz
I did notice that by applying Chris’s plugin it added extra sound to the blank space at the end of the track. It sounded like loud white noise. Not sure why that happened. I just removed it and replaced it with the blank space that was there before I applied compression.
I just submitted the files for review. We’ll see what they say …
I think it’s a minor defect of that compressor that it can “bloat” the start or end of the selection.
Gale
Chris is a “look ahead” compressor. Beginnings and endings can confuse it, as when the show suddenly drops off a cliff at the end. Remember, this is a home-grown Audacity plugin, not a commercial product.
Koz
I just submitted the files for review. We’ll see what they say …
We are so on the edges of our seats — across nine time zones.
While we’re waiting, I wouldn’t complain if you posted some of the Before performance. I want to get a feel for the starting point. Chris won’t fix you if you’re just too far off. I want a healthy chunk of reading, not the second or so that the forum will accept. Let me know and I’ll accept it through email and post it on my web site. It will be publicly visible to search engines up there, so try for a generic reading. Of course, it will be visible on the forum, too just not that much of it.
Did you conform to all the guidelines? Mono track with 128 or higher MP3 quality?
Koz
The built-in Compressor also employs “look ahead” without suffering from that exact problem.
Gale