Normalization
Also known as: normalisation, level matching
Adjusting a track so its level matches a target, either by its loudest peak or by how loud it sounds across the whole duration.
Updated
What it means
Normalization applies one gain change to bring a recording to a target. Nothing inside it changes relative to anything else, which distinguishes it from compression, where the quiet and loud parts move differently.
There are two kinds and they answer different questions. Peak normalization sets the loudest Audio peak to a chosen ceiling, while loudness normalization sets the perceived level measured in LUFS across the whole file.
The second is almost always what is wanted for video. A track normalised by peak can still sound quiet if it has one sharp transient and nothing else, which is a common outcome on action audio.
How it works in practice
Normalising each clip before assembly removes the worst inconsistency. Footage from several cameras and several days arrives at wildly different levels, and matching them first is what makes the balance of a finished mix tractable at all.
A ceiling below maximum is the sensible target. Leaving a few decibels of headroom means later stages can add music or an effect without pushing the result into Clipping.
For the finished video the target is whatever platforms expect, so the upload is neither turned down nor raised on arrival, and the balance chosen in the mix survives into what a viewer actually hears.
What to watch out for
Normalization cannot fix dynamics. A clip whose speech is inaudible under wind stays inaudible after the whole thing is made louder, because the relationship between the two is unchanged.
Peak normalising a clipped recording also achieves nothing useful. The flattened tops are brought to the ceiling precisely, and the distortion that was already baked in comes along with them.
Normalising to maximum loudness is counterproductive once platforms are involved, since the excess is taken straight back off on upload while the squashed Audio levels remain. The result is a flat track played at the same volume as a dynamic one.
How RawClip handles it
Clips are level matched before they are assembled, so a highlight cut from several cameras and several days does not jump in volume between shots. The finished mix is then set to the loudness platforms expect rather than to maximum.