Audio Ducking
Also known as: Auto Ducking
What Is Audio Ducking?
Audio ducking lowers one sound while another one plays, then brings it back up when the other sound stops. The usual case is a music bed under a voice: the music dips while someone talks so every word stays clear, and it rises again in the gaps between lines. Radio presenters, podcasts and narrated videos all use it, and radio dubbing uses it to lower the original speaker under a translator’s voice.
A ducker does it on its own. It listens to the voice, and when the voice gets louder than a threshold, it turns the music down by a set amount until the voice stops. Editors without one get the same result by hand, with keyframes on the music’s volume line.
How Much Should You Lower Music Under a Voice?
Two published starting points sit 10.5 to 12 dB below the music’s full level: Audacity’s Auto Duck lowers music by 12 dB by default, and the 30% volume the CE.SDK audio docs suggest under a voiceover measured 10.5 dB down in our test. For that test, we exported a 1 kHz tone from CE.SDK 1.85.1 at several volume settings and measured each file.
-
Set the music to 25 to 30% volume
A setting of 25% measured 12 dB below full level, the same dip as Audacity's default, and 30% is the level the CE.SDK docs suggest. Treat either as a first setting for the voice you have.
-
Halve the volume for every further 6 dB down
The setting is a plain multiplier on the signal, so 50% measured 6 dB down and 10% measured 20 dB down. For any other value, 20 × log10 of the setting gives the change in decibels.
-
Stop once every word is clear
Too little reduction lets the music cover the voice. Too much makes the dip itself easy to hear each time the voice starts or stops, which Sound On Sound and Boris FX both warn about.
What Do the Settings on an Audio Ducker Do?
Every ducker works with two tracks: the one it lowers, usually music, and the one that triggers it, usually the voice. Logic Pro feeds the voice in through a side chain, and Audacity listens to the first unselected track below the music. The amount sets how far the music drops, and four settings decide when and how fast.
The threshold decides when the music drops
The ducker waits until the voice gets louder than this level, so breaths and room noise between lines leave the music alone. Audacity sets it at -30 dB by default, and Corel VideoStudio calls the same control Sensitivity.
Attack sets how fast it goes down
A short attack drops the music almost at once, and a long one fades it down gently. Audacity’s Fade Down takes 0.5 seconds, and duckers that read the recording ahead, as Logic Pro’s does, can start the fade before the first word.
Hold keeps it down through short pauses
Without a hold, the music would rise in every gap between words and fall again, a fast up and down that Logic Pro’s manual calls chattering. Audacity’s Maximum Pause keeps the music down through any pause shorter than 1 second by default.
Release sets how fast it comes back
Once the voice has finished, the music climbs back to its full level over the release time. A slow release lets it swell back in, which suits the end of a narration, and Corel VideoStudio calls this setting Decay.
When Should You Duck Audio?
Duck when a voice and music play together and the voice carries the message, as in a narrated product video that lowers its music under every line. When someone talks from the first second to the last, one lower music level does the same job with no ducking at all.
| Use | Why | |
|---|---|---|
| A narrator speaks over music, with gaps between lines | An auto ducker keyed to the voice | The music drops under each line and fills the gaps on its own |
| Someone talks for the whole video | One lower level for the whole music clip | With no gaps to fill, a fixed level keeps the voice clear |
| Two or three lines over a long song | Volume keyframes set by hand | A few dips are quick to draw, and you pick the depth and length of each |
| An interview dubbed into another language | A ducker on the original voice | The speaker drops under the translation but can still be heard |
| A dance track where the kick drum should punch through | Sidechain compression keyed to the kick | The bass dips on every beat, which gives dance music its pumping sound |
| A J cut, where the next voice starts under the last shot | A short dip in the last shot's sound | Both sounds play for a moment, and the new voice has to come through |
| A music-only intro or outro | No ducking, the music at full level | Nothing needs the room, so a dip would only make the music quieter |
How Clip Volume Works in CE.SDK
Every audio clip in CE.SDK video editing, and the sound of every video clip, has a volume of its own, so a music bed can sit below a recorded voiceover on the same timeline. Your users set an audio clip’s level with the volume control on the selected clip, and the engine API sets the same levels from your code, for example when a template places music under a voice.
Each clip keeps its own volume
Music, voice and video clips each run from 0 to 100%, and muting a clip keeps its level for when it is unmuted. On the web, Separate Audio moves a video clip’s sound onto audio clips of its own at the same volume and mutes the video, so that sound can be trimmed and leveled apart from the picture.
Fades ramp a clip in and out
Your code sets a fade at a clip’s start or end with a length in seconds and an easing curve, and the gain ramps between silence and the clip’s own volume. The fade follows the clip’s edge through a trim, and a split keeps only the two outer fades.
A voice and its music add up
CE.SDK adds overlapping clips together at the levels you set. In our test, two tones that each peaked at 60% of full scale went more than 1 dB over full scale in both the WAV and the AAC export, so the music needs to sit below the voice.
FAQs
Sidechain compression is one way to duck. A compressor on the music listens to the voice through its side chain and turns the music down by more the louder the voice gets, while a dedicated ducker, such as the one in Logic Pro, drops it by one set amount. Dance producers also key a compressor to the kick drum on purpose, which gives their tracks a pumping sound.
No. A fade takes a clip down to silence, usually at its start or end, while ducking lowers the music to a set level in the middle of a clip and brings it back. In CE.SDK a fade works on top of the clip volume, so a clip set to 30% fades in from silence to 30%.
Yes. Final Cut Pro's Range Selection tool lets you mark a stretch of the music and drag its volume line down, and the keyframes are created for you. Apple's guide suggests this for ducking. Corel VideoStudio's Audio Ducking command also leaves keyframe nodes on the volume line, which you can drag, add or remove afterwards.
It goes down and up too often or too fast. A short release lets the music jump back between words, so duckers have a hold setting (Maximum Pause in Audacity) that keeps it down through short gaps, and a smaller amount makes each change harder to hear.
Links
Documentation and references for this concept.
Video Editor SDK
Let your users balance music, voiceover and video sound clip by clip on one timeline.
Adjust Audio Volume Guide
Set and read a clip level from 0 to 1, mute it, and balance music against a voiceover.
Fade Audio In and Out Guide
Ramp a clip up at its start and down at its end, with an easing curve.
Related Terms
Timeline
The time-based editing interface in CE.SDK's video editor where users sequence, trim, and synchronize all media elements…
B-Roll
Supporting footage, such as cutaways, inserts and establishing shots, cut in over the main footage of a video to show wh…
Video Transitions
What happens on screen where one clip ends and the next begins: a plain cut, or an effect such as a cross-fade, fade, wi…