Normalization is another in a long list of audio recording concept whose term is both misleading, needlessly “tekky” sounding and confusing. Basically it means turning up the volume (technically “gain” – but that’s another story) of a piece of audio so that the loudest/tallest peak is as loud as possible without clipping. What’s so hard about that? It’s usually done to increase the volume of your audio to as loud as it can be, all other things being equal (no other editing done to the audio). But the term “normalization” implies that by doing this to multiple files, they’ll all be the same loudness – at least that is what normalizing something means to me. But surely you can see that this simply is not true. The amount that audio is turned up completely depends on how high the highest peak is. Say you have two recordings of equal average loudness. If one recording had a single very loud peak, a quick shout, say, and left only about 3 dB while normalizing before it reached max volume, that entire file would only be turned up by 3 dB. If another recording’s loudest peak were not very high, allowing, say 10 dB of available volume increase during normalization, that file’s volume will be increased by 10 dB. See? These files, who had equivalent average loudness prior to normalization now have radically different volumes.
That (the idea that after normalizing multiple audio files, they’ll all be the same loudness) is only one of the normalization myths discussed in this article: