This effect does not change the dynamics, it only changes the overall level.
Recordings for commercial use (such as audiobooks) are frequently required to meet a specified “dB RMS” range. For example, the requirement specified by ACX / Amazon audiobooks is -23 to -18 dB rms.
The RMS level is a better approximation of “loudness” than the peak level, so this effect may be useful for amplifying tracks / audio clips to approximately the same level (though the ReplayGain plug-in may be better for this purpose).
This effect may be used on mono or stereo tracks up to 200 million samples duration (about 1 hour 15 minutes at 44100 Hz sample rate - see warning below).
The RMS for stereo tracks is calculated as the “square root of (the average of the squared amplitude for both tracks)”. Reference: http://replaygain.hydrogenaud.io/proposal/rms_energy.html
Expected accuracy is within 0.1 dB.
Like many Nyquist Plug-ins, the maximum selection is limited by available ram (up to a maximum of 2 GB). The selection length is therefore limited to 200,000,000 samples (about 1 hour 15 minutes for a 44100 Hz sample rate track. An error message is returned if the selection is longer than 200 million samples. If you wish to apply this effect to very long recordings, it may be applied to sections of less than 1 hour.
On long selections the effect is slow, taking several minutes to process a 1 hour track.
If the computer has less than 2 GB of physical ram free, Audacity will crash if the selection is larger than the available ram.
The RMS level of a signal is always* lower than the peak level, so to avoid clipping the normalization level must be substantially less than 0 dB. The default is -20 dB, which should be safe for most recordings.
(* Except for square waves, where the peak level and RMS level are the same)
rmsnormalize.ny (2.51 KB)
UPDATE: For Audacity 2.1.3 or later, use the updated version here: https://forum.audacityteam.org/viewtopic.php?p=325078#p325078