(Question is on the third paragraph, if you don’t care for backstory) I was given a job to record an audio book on ACX, which I loved doing (hated editing, but everything has a drawback).
I finally finished the book, and start doing the final edits, only to find that the RMS is not inside the desired parameters for ACX (not surprising, I acted out the lines as they were read, so whispers were softer, shouts were louder, etc).
Knowing nothing about what this was, searches said to normalize the volume -but when I do, all the lip smacks, breaths, tongue clicks, and all the other unwanted noise I worked to reduce beyond audible levels are raised to be main stars of the audio file, and many of the voices I worked so hard to get sounding good, now sound like they are being read into milk jugs.
Is there a way to either use normalize without raising the unwanted sounds and trashing the louder voices, or to reduce the peak volume of the entire set beyond the 10db limit of the “hard limiter” (which would allow me to normalize over a smaller range and thus maybe keep the unwanted sounds quieter)?
If not… then I’m not sure what to do… because this is a “quality assuring requirement” (that destroys the quality) for ACX to allow a file to be uploaded.
There is a link to the new ACX mastering macro here (for Audacity 3.6 or newer). It will nail the peak & RMS requirements.
Any regular-linear volume adjustments change the peak, RMS, and noise levels by the same amount and usually either the peaks or RMS level end-up out of-of-spec.
The macro adjusts the RMS level first (a liner adjustment), then a limiter is applied to push-down the peaks if necessary (nonlinear, with almost no effect on the RMS levels).
Typically, the RMS levels end-up being boosted, which means the overall loudness is boosted, including any background noise (or other noises) and the noise floor sometimes goes out-of-spec. If that happens it has to be delt with separately.
I don’t know what MBR is… The ACX requirements say that the peaks can’t exceed -3dB, the RMS level needs to be between -18 and -23dB, and the noise floor has to be -60dB or better. But if there is no noise (minus infinity) your audio will be rejected for “over processing”.
This is why it is recommended that you wear large, good-quality headphones listening to yourself as you perform. Loud-excited voices will be painful to hear and quiet, expressive voices will vanish. A little experience and rehearsing with the headphones can produce a perfect electronic-theatrical performance.
Fixing a fully theatrical performance in post production editing without breaking anything can be difficult as you’re finding.
A note about Mastering. It does not do full Compressor/Limiter tricks. It will not go through the performance jacking the volume up and down like a broadcast station.
I wrote this simplified instruction a while ago.
Record your chapter in Mono (not Stereo) in a quiet, echo-free room. Include two seconds of hold-your-breath room tone at each end. Export a Protection WAV file—errors and all.
Correct the accidents, fluffs, stutters, mis-speaks, and pronunciation errors. Make it sound pleasant. Use “Room Tone” for editing and not generated silence .
Apply 36Audiobook-Mastering-Macro. Check the chapter with ACX-Check. Export your Edit Master WAV file. Export your 192 Constant-Quality, MP3 file for submission. Go make dinner.
That doesn’t seem too dreadful, does it? I can do a very entertaining soft-shoe at each one of those steps, but I believe that’s the whole world.
I think I mentioned somewhere at the top that my test submission (back with they allowed such things) was Practically Perfect in Every Way, but I got killed with mouth noises.
