Sidechain Ducking

In plain words

The music gets quieter automatically only while the narration plays, so the words don't get buried.

When to use it

Every ad that lays music under narration or dialogue.

Example videos

▶

▶

In AI video, like this

Level each narration line to the same volume (-18 LUFS) and lay them on one track, then split that track with asplit and use one copy as the ducking signal. [music][vo1]sidechaincompress=threshold=0.05:ratio=4:attack=40:release=500:makeup=1[duck] → [duck][vo2]amix → loudnorm. The lower the threshold and the higher the ratio, the more the music drops. If you lower the music itself first (0.55x) and compress lightly, it is less noticeable.

One-click prompt

Lower the music automatically while the narration speaks, using a gentle setting (threshold 0.05, ratio 4) so the music does not pump, and normalize each narration line to the same loudness first: ffmpeg -i music.mp3 -i vo.wav -filter_complex "[1:a]asplit[sc][vo];[0:a][sc]sidechaincompress=threshold=0.05:ratio=4:attack=40:release=500[duck];[duck][vo]amix=inputs=2:duration=longest:normalize=0[a]" -map "[a]" out.wav

See it in real films and ads

Wikipedia — Ducking

AI-made examples

Not verified

Recommended for scenes like these

Techniques that pair well