After a lot of experimenting over many years, I am at a loss for how I can achieve the following (without lots of "painful experimentation"):

* normalize the RMS volume of a group of samples
* max value of the loudest peak in the group doesn't exceed xxx db
* no limiting/clipping/compression

Imagine a (non-existent) command line such as:

$ normalize-audio --rms --max-peak=xxxdb --as-lossless-as-possible

note that with normalize-audio 0.7.7:

$ normalize-audio --peak --no-adjust
$ normalize-audio --no-adjust

result in exactly the same output:

level peak gain

-44.9045dBFS -26.6550dBFS 32.9045dB china-22in_zildjian-1.wav

-36.0490dBFS -18.3966dBFS 24.0490dB china-22in_zildjian-2.wav

-33.1798dBFS -14.9723dBFS 21.1798dB china-22in_zildjian-3.wav

-25.6555dBFS -5.5081dBFS 13.6555dB china-22in_zildjian-4.wav

-18.4676dBFS -1.4021dBFS 6.4676dB china-22in_zildjian-5.wav

-17.8024dBFS -0.0540dBFS 5.8024dB china-22in_zildjian-6.wav

In this case, the file with a peak of -0.054db is adjusted by +5.8db in both cases, resulting in clipping/limiting etc.

Shane Richards

Producer, Composer, Multi-instrumentalist

Josh Music

shanerich@email.com