The words you are searching are inside this book. To get more targeted content, please make full-text search by clicking here.
Discover the best professional documents and content resources in AnyFlip Document Base.
Search
Published by vdovitsky, 2019-01-02 17:34:43

mastering for ACX with Audacity

mastering for ACX with Audacity

Audio Mastering and Cleaning procedure
in Audacity to meet the official
ACX/Audible requirements

Preamble

ACX/Audible as a one of the major platforms have written their own technical protocol (can be found
here https://tinyurl.com/ybemxc7p), the main reason is to provide listeners with more-less qualitative
and unified (in technical terms) content.

Why they have set the loudness corridor (-23…-18 dB and peaks at -3 dB)? They want to give the
consistent content in terms of volume, so user doesn’t have to turn the volume knob while listening to
the content from different producers.

Why they’ve set the Noise Floor? Well, it’s obvious… noise is very distracting and should be removed in
any way. The important thing here is not to get new digital artifacts (so called ‘musical noise’ or
‘robotic’, or ‘under-water-noise’, etc).

We’ll try to get it done with Audacity — https://www.audacityteam.org/

Preparation

1. Please go download and install the Audacity;
2. Please download these two files (we are going to use them extensively) —

https://www.dropbox.com/sh/td8ltezg7vvzhlx/AABHFKfW7lhv0yaqpOw0jSrda?dl=0
3. While ‘SetRMS’ file can be stored anywhere on your computer, the ‘ACX check’ should be placed

into the Plugins folder of the program (windows path — C:\Program Files (x86)\Audacity\Plug-
Ins);
4. Open up the Audacity. Go to menu Effect -> Add Remove Plugins -> ACX check and hit enable.
Then it should be available under the Analyze menu.

We are ready to start

1. Load the project or audio file itself to the program.
2. First thing — if you see something like this (picture), be aware of high volume and possible

distortion.

3. Don’t play it and decrease the amplitude to the more safe volume. Effect -> Amplify -> Set
peaks to -6 dB

4. Listen to it first time (at least episodically) and notice any kind of permanent noise, distortions,
cracks, crackles and such. If sound doesn’t seem to be defective (there’s no terrifying level of
permanent noise — when you cannot even hear the voice, there’s no excessive level of
reverberation or distorted sounds), we can proceed.

Equalization

1. First of all we have to get rid of any low rumbling sound (usually electricity hum or thumps
transferred by construction). Go to Effects -> High Pass Filter and apply it at 100 Hz with 24 dB
per octave slope.

NOTE — since we deal with the human’s voice, the fundamental tone area starts from around
100…120 Hz for male voice and 160…200 Hz female (very roughly and depends). If you think the
100 Hz filter cuts something significant in the voice (‘body’, so to speak) — it’s usually for
baritone and bass male vox, — try to set the frequency a little lower (usually 60…80 Hz is pretty
sufficient).
2. Go to Analyze -> Plot Spectrum and get the idea about the frequency response of your voice.

a) The first significant ‘hill’ on this picture at around 160 Hz is the fundamental area — ‘body of the
voice’;

b) Next hill (area in between 200…400) usually refers to most annoying ‘standing waves’ of the
room. By decreasing this area you’ll be able to reduce the ‘roomy’ sound. However, not
completely…

c) Next area 400…1,000 usually refers to ‘nasal’ or ‘telephony’ tones;
d) Clarity and brightness usually sit in between 2k and say 4k (again, roughly);
e) And the sibilance area is around 4k to 8k — refers to intelligibility;

3. Our goal is to get as clear and natural voice as possible. As a benchmark for the frequency
response just keep in mind that the difference between the fundamental tone and sibilant area
should be at around 30…35 dB (see the left ruler at the plot).

Now,

a) If it seems like the voice sounds a bit ‘muddy’ — try to slightly reduce the fundamental tone and
increase 2…4k (or even higher, just listen to it critically), get the right and natural balance;

b) If it feels ‘roomy’ — work with 200…400 Hz (not exactly though, try the lower and higher as
well);

c) Feels too ‘nasal’ — now you know which area responsible for that.

You got the idea. Listen to the sound (your headphones/speakers must be of good quality) and compare
your feelings with what you see on the spectrum graph. Again, our goal is the tendency with the slope
30…35 dB from lowest area of the voice to the sibilant without significant ‘hills’ (they usually tell us
about too colored sound).

4. Go to Effect -> Equalization and reflect what you found (use Graphic mode, it’s easier to work
with). If needed, play with it for a while to get better results. Sure thing, you can save this curve
as a preset for this particular voice to use it in the future.

Noise Reduction

To ‘teach’ your noise reduction module properly you need to pick the most steady sample as possible.
Choose the ‘spectrogram’ mode from the dropdown menu and pick the most consistent noise sample
(without vertical spikes):

1. Go to Effect -> Noise Reduction and get a profile.
2. The next step is very important and probably will need to be repeated until satisfactory result.

Adjust the parameters in a right way. As a start point you can choose the following:

Useful articles:

https://www.vo2gogo.com/an-extremely-dangerous-article-about-noise-reduction-in-audacity/
https://manual.audacityteam.org/man/noise_reduction.html

3. Play around and try to get a compromise between the noise floor and new artifacts. It’s very
important! New digital artifacts are very annoying and are going to be considered at QC stage as
a defect. And that is true.

Normalization

Now it’s time to get the right loudness. It’s easy to do.
1. Open up the ‘SetRMS’ file and copy the whole thing to the clipboard;
2. Go to Tools -> Nyquist Prompt (in earlier versions it was under the Effects menu) and paste the
code;
3. Apply with target RMS level -20 dB (actually in the middle of the target volume, right?) and…
wait… I couldn’t notice any of notifications, so just wait.
4. Once done, go to Effect -> Limiter and apply with these settings:

Why -3.5 dB (ACX asks for -3.0). This is an insurance measure. After the conversion to MP3, you can get
some spikes (+0.1 … +0.3 dB) and an ACX robot is going to refuse it.

Final Steps

1. Go to Analyze -> ACX check and see what you get:

Note — it might be that the noise floor still is fairly high. Just apply the noise reduction once more, but
with slightly reduced strength (say 5 to 10), should be okay:

2. Check again.
3. Edit your stuff (re-takes, breaths, clicks and so forth).
4. Check again with ACX check plugin (just to be sure).
5. Render as MP3. File -> Export -> Export as MP3


Click to View FlipBook Version