Glossary
What is audio ducking?
By the KeyMo team · Updated
Audio ducking automatically lowers one sound, usually music, while another sound, usually a voice, is playing, then lets it return to normal when the voice stops. The result is that speech stays clear over a music bed without you riding the volume by hand. It is the audio version of stepping aside.
How ducking works
A ducker listens to one track, the trigger, and turns down another track while the trigger is loud. Four settings do most of the work. The threshold decides how loud the voice must be to trigger it. The amount is how far the music drops. Attack is how fast it falls once speech starts. Release is how long it takes to come back. Editors do the same job by hand with volume keyframes, and some tools do it automatically.
| Setting | Too little | Too much |
|---|---|---|
| Amount | Voice is still masked | Music vanishes and the clip feels empty |
| Attack | First word is clipped by music | Music dips before anyone speaks |
| Release | Music snaps back mid-sentence | Music stays low through the pause |
An example with a guest answer
Picture a 30-second clip with a quiet instrumental bed. The guest speaks for 22 seconds, pauses for two, then finishes. With no ducking, the music sits at one level and the guest's consonants get lost on a phone speaker. With ducking, the music drops a few decibels under the first word, stays under the whole answer and rises in the pause, so the pause feels intentional. Nothing else about the edit changes.
Test it the way a viewer hears it. Play the clip on a phone at normal volume. If you have to lean in to catch any word, duck deeper or lower the bed overall.
Where it matters for clips
Spoken-word clips from podcasts are mostly voice, so music is a mood layer and must never compete. Instagram says a reel can carry up to 20 audio tracks, so voice, music and sound effects are expected to be layered. EBU R 128 adds that peak normalisation has led to large loudness differences between programmes, and recommends loudness-based normalisation measured over the whole signal. Keep your voice level consistent from clip to clip. See LUFS for the measure.
Common mistakes
- Setting the music so loud that deep ducking is needed. Start with a quiet bed.
- Fast release on a track with long breaths, which makes the music pump.
- Using music with vocals under spoken words. Two voices fight even when ducked.
- Skipping the check on a phone speaker.
Where KeyMo touches it
KeyMo has a music library organised by mood that you can add to a clip. This page does not claim that KeyMo ducks music automatically, so play the finished clip and listen to the loudest spoken line before you post.
Sources: Instagram for Creators: Level Up with Instagram Reels (checked 11 October 2026); EBU R 128: Loudness normalisation and permitted maximum level of audio signals (checked 11 October 2026).