ACX Check Fails for Whole Chapter But Not Individual Parts, Unless Done Individually

[GJ_Herman]

I’ve recorded a chapter for my audiobook and have followed the ACX Audiobook Mastering instructions exactly. (I also experimented after the fact with Noise Reduction, but the issue below is without that step.)

I ran the steps on the whole track (‘select’ button next to up arrow on track), and am running version 3.2.5 on a Mac M1 (the fan never spins up). When I run ACX Check on the whole track, I fail the RMS level (too quiet). Peak is -3.48 (pass), RMS is -25.08 (Fail, too quiet), and Noise floor is -66.92 (pass).

When I select individual elements in that track and run ACX Check, each and every slice of the puzzle passes. As I ran the individual steps on the whole track and all its elements, I would expect that all would either pass together or fail together, but obviously my understanding of audio mastering is erroneous.

That’s only when applying the 3 steps of the Audacity Audiobook Mastering instructions on the whole track all at once. Each individual part will then pass while the whole fails. When I individually process each of the parts, and then run ACX Check on the whole track, the whole track and each individual part passes. But, I think that defeats the purpose, whole chapters (and maybe even later the whole book) need to be processed all together for a common, consistent sound.

I’ve recorded an audio sample, a la Koz’s suggestion (I’m hoping he’s still around) in the hopes that someone could point me in the right direction, but I see no option to upload it here to this post.

I have a quiet room (not dedicated), a Shure MV7 with Mic Gain set to 36db (max, although I’ve tried everything from 0-max), an t.bone Micscreen, pop filter, Mac recording it is set a distance away (although it makes no sound), and reading from a Kindle or Macbook Air M2 (no fan installed) on a separate electrical circuit. All other electronic devices are turned off and unplugged.

I’m concerned that the individual parts of the chapter pass but the whole chapter doesn’t, or I have to process each part individually and then end up with a disparate, disjointed mess come final processing.

Any suggestions? Is it normal to process each part separately? (Filter Curve EQ, Loudness Normalization, Limiter)

Thanks in advance.

GJ

[steve]

@GJ_Herman

I ran the steps on the whole track … When I run ACX Check on the whole track, I fail the RMS level (too quiet). … RMS is -25.08 (Fail, too quiet),

@GJ_Herman

When I individually process each of the parts, and then run ACX Check on the whole track, the whole track … passes.

What are you doing different in those two cases? What are the “processes” that you are applying? Are the “steps” in the first case the same as the “process” in the second case?

[GJ_Herman]

Hi Steve,

Sorry, I don’t mean to be confusing.

In every process, I run the following steps:

• Effect > Filter curve… > Manage > Factory Presets > : Low roll-off for speech > OK.
• Effect > Loudness Normalization…: Normalize RMS to -20dB > OK.
• Effect > Limiter: Soft Limit, 0.00, 0.00, -3.50dB, 10.00, No > OK.

I do this either for a) The whole track (chapter) all at once or b) each “element” (not sure what to call them; paragraphs/scenes?) of the track individually. Same steps/process for each attempt.

I don’t know if doing each paragraph/section/scene separately and then combining them together as an end result will be different than running the whole chapter at the same time and together, but I certainly don’t want to think I’ve succeeded in passing what needs to be passed and just find out later that I have to re-narrate the whole book again.

Thanks for your reply and in advance for any help or clarification you can provide.

Regards,
Greg

[steve]

The Filter Curve effect will be the same whether you apply it to the entire recording in one go, or section by section.

The Loudness Normalize and Limiter effects may give different results when applied to the entire recording to the result of applying section by section. To explain why that is, consider this example (below). Both “Part 1” and “Part 2” are pink noise, which represents the recording. Part 1 has an RMS level of around -15 dB, and Part 2 has an RMS level around -27 dB.

If I Normalize the loudness of the entire track to -20 dB, then the result will be as shown below. The average level of the entire track is -20 dB. Part 1 has an RMS level of -17.26 dB and Part 2 has an RMS level of -29.26 dB.

If on the other hand I Loudness Normalize Part 1 to -20 dB, and then Loudness Normalize Part 2 to -20 dB, each part will have an RMS level of -20 dB:

In an ideal world, each time you record a chapter of the book, you would be using the same microphone, in the same room, with the microphone in exactly the same position, speaking from the same position, speaking at the same volume, with the same amount of energy and a consistent tone … In short, the recording will be consistent throughout the entire book.

You should aim to be as close to that ideal as possible, but in real life it is likely that a recording on one day may be a bit different than on another day. To counter such differences, it becomes necessary to process each “different” part separately so as to make the sound consistent throughout the book. If the recordings are too different, it may not be possible to make them sound consistent, in which case you will need to re-record one or more parts.

[GJ_Herman]

Thanks a lot for your detailed and clear explanation. It makes so much sense when you explain it, that I wonder why I didn’t figure it out for myself.

Each of those segments was recorded right after each other, because, for some reason, I can’t seem to read a whole chapter without mispronouncing something. I’m surprised that so much variation can occur even when only minutes separate each recorded segement. But, it makes sense, as there’s always variation in voice due to frustration regarding having to narrate something ten times before it’s good.

Thanks again!