Sidechain compression
Also known as: sidechaining, pumping compression
Using one sound to control the level of another automatically, most often letting a kick drum or a voice push the rest of the mix down.
Updated
What it means
Sidechain compression takes its control signal from somewhere other than the track it is processing. A compressor on the music listens to the speech, so whenever words appear the music ducks without anybody touching a fader.
That is the mechanism behind every automatic level change in a finished video. The audible result of Audio ducking under a voice is sidechaining doing the work, reacting within milliseconds and recovering afterwards.
In music production the same trick produces the pumping effect under a kick drum, where the bass drops on every Downbeat and the groove comes from the gap. It is the same process aimed at a different control signal entirely.
How it works in practice
Attack and release settings decide whether it sounds deliberate or broken. A fast attack catches the first syllable, while a release too fast makes the music jump back up between words and breathe audibly.
The amount of reduction matters less than its consistency. A few decibels applied smoothly is invisible, where a large reduction applied unevenly draws attention to the processing rather than to the speech.
The control signal usually wants filtering before it is used, so that Wind noise on the camera track does not trigger ducking when nobody is actually speaking at all.
What to watch out for
A release that fights the music is the common fault. Ducking that recovers in the middle of a bar pumps against the Beat grid, and the mix sounds unstable rather than attentive.
Triggering from an unfiltered track is the other. Gusts, bumps and engine noise all look like events to a compressor, so the music ducks repeatedly for sounds no viewer is trying to hear.
Too much reduction also leaves holes. Dropping the music almost to nothing under a single word makes the gap more noticeable than the word, which defeats the purpose of ducking it at all.
How RawClip handles it
The music level is driven by the speech in the camera audio rather than by a fixed curve, so it steps back when there is something to hear and comes forward when there is not. The control signal is filtered first so wind does not trigger it.