Recall that audio, in a computer, can have intensity anywhere along a scale of number values - and that corresponds to volume.  We could think of that scale as running from 0% to 100% (and Soundtrack Pro does just that by default), but, since we're video editors, we're more likely to think of the scale like Final Cut does - in terms of decibels.  Decibels run from -96dB (or sometimes "-Infinity dB") up to 0dB - with 0dB being the loudest possible sound that the computer can represent.

Clips can have audio levels in any range within the volume scale.  One clip can be very quiet, and the loudest parts could only get up to, say, 40% or -24dB.  The very next one could have loud parts that get all the way up to the 98% mark, or just under 0dB.  And if you crank up your TV or computer speakers to hear the quiet clip, you'll be blown away when the loud one comes on.

It would be in our interests to have all of our audio in basically the same volume range, but which range should we choose?  We could try and make most of our video quiet, so that the BIG LOUD EXPLOSIONS have more volume range to get REALLY REALLY LOUD.  But recall the diagram comparing 8bit to 16bit audio:

Essentially, if we were to only use a little bit of our volume range, we'd be sacrificing detail in just the same way as the left-hand side of the diagram does.  Out of that ethos, most broadcasters tend towards a relatively loud average volume.  For broadcast purposes, we'd like for our audio to tend towards the range from -6dB to -12dB.  For multimedia and web delivery, it's safe to aim closer to the 0dB mark - but be careful not to actually hit 0dB, or your audio will clip and sound bad.


Normalization is simply the process of making sure that our audio is in this relatively high range - and that it's consistent between clips, or even between parts of clips.