DaVinci Resolve Audio Ducking: Ducker, AI & Sidechain

Quick answer: put dialogue and music on separate tracks, select the music track, switch on Ducker and pick the dialogue track as its Source. Start with a modest Duck Level in the Inspector, then tune Lookahead, Rise Time, Hold and Recovery until the speech stays clear and the music doesn't audibly pump.

When a track or bus should control another track or bus, use the current Fairlight Dynamics sidechain instead. Manual keyframes are for the editorial exceptions.

Audio ducking has one job: make room for the words while the music keeps doing its emotional work. The setup takes a few minutes. The hard part is keeping the movement subtle enough that viewers hear the dialogue, not the mixer at work.

DaVinci Resolve 21 gives you three tools that tutorials often lump together as one feature. Ducker Track FX is the fastest direct method. Fairlight sidechain compression adds routing and full compressor control for tracks or busses. In Studio, the AI Audio Assistant can apply ducking as part of a much broader automatic mix. Pick one on purpose instead of stacking all three; for a single voice over a music bed, I'd start with Ducker.

Pick the Right Ducking Method

Ducker Track FX works in the free version as well as Studio. The Ducker Track FX section of the Resolve 21 manual documents it without the Studio Version Only tag that Blackmagic puts on paid features such as the AI Audio Assistant.

The 21.1 edition of the manual still lists it that way. So on Free you don't get the AI whole-mix pass, but the ducking itself is yours.

Prepare Dialogue and Music Before Ducking

Automation can't fix a timeline where everything is mixed together. Dialogue goes on dialogue tracks, music on music tracks. Drop a voiceover clip between music clips on one track and a track-level Ducker has no way of knowing what you meant.

1. Move speech onto clearly named tracks such as Dialogue or VO.
2. Keep the music bed on its own Music track.
3. Set a sensible static balance first. The music should already sit close to where you want it when nobody's talking.
4. Clean obvious bumps, clothing rustle and long empty stretches out of the dialogue source.
5. Listen through once with no ducking and mark the places where the voice actually gets masked.

Ducking reacts to signal level, not meaning. A breath, a mic bump or loud room tone can trigger it just like a word can. Basic clip cleanup from the mute and split audio workflow often prevents more false triggers than another round of fiddling with the controls.

Set Up Ducker Track FX

1. Open the Edit or Fairlight page and show the audio mixer.
2. Select the music track. That's the target, the track whose level goes down.
3. Open Track FX from the track's Inspector or its mixer channel and switch on Ducker. The mixer only shows Voice Isolation and Dialogue Leveler by default, so if Ducker isn't there, click the three dots at the upper right of the Mixer and pick it from the Visible Track FX submenu.
4. In Source, choose the dialogue or voiceover track that should trigger the dip.
5. Several speakers on separate tracks? Add them as extra sources with the plus control.
6. Loop a section with speech, a silence and another phrase while you tune the controls.

Put the Ducker on the music, not on the voice. The source is only the detector; the target is the signal whose level changes. Swapping source and target is the classic setup mistake.

Blackmagic's latest Fairlight overview describes the same flow: open Ducker on the target track, pick one or more dialogue sources, then set the depth and recovery. In the full Ducker window, the manual has you Command-click the extra track names; on Windows and Linux that's Ctrl-click, following the manual's general shortcut rule. In the Inspector, the plus control next to Source 1 creates Source 2 and Source 3.

If you'd rather watch the clicks happen, Dunna Did It's January 2026 ducking walkthrough shows the same setup on screen in Resolve Studio 20, and the controls it uses (Source, Duck Level, Lookahead, Rise Time, Hold, Recovery) carry the same names in the current Resolve 21 manual.

Understand the Ducker Controls

• Duck Amount / Duck Level: how far the target track drops while a source is active. The full Ducker window calls it Duck Amount; the Inspector calls it Duck Level.
• Lookahead: lets the dip begin before the detected speech is audible, so the first syllable doesn't have to fight the music.
• Rise Time: how long the Ducker takes to reach the full reduction you've set.
• Hold: how long the target stays down, which carries it through a short pause.
• Recovery: how quickly the music comes back to its original level after the speech.

Set depth first, then timing. Raise Duck Amount until the words are clear, but stop before the music sounds like it fell through a trapdoor. Next, tune the entrance around the first syllable. Last, set Hold and Recovery so short pauses inside a sentence don't make the music breathe between every few words.

There's no universal dB preset. The Resolve 21 manual sets 2.7 dB as the default and says a value between 2 and 5 dB works best most of the time. That's a starting range, not a delivery standard.

A sparse piano bed, a dense rock track, a quiet narrator and an energetic presenter all mask each other differently. Tune on the actual program and monitor at a realistic playback level.

Use Fairlight Sidechain Compression for Tracks or Busses

Ducker is a Track FX. When several dialogue tracks feed a dialogue bus and the music stems feed a music bus, put the Dynamics sidechain on the music bus instead.

In Resolve 21 you choose the source track or bus directly inside the Sidechain module. The manual calls that section redesigned: it replaces the older method of feeding the sidechain from a Send. So if a tutorial tells you to enable Send on the dialogue channel and Listen on the music, it's showing the old routing. The Fairlight bus routing guide explains the submix structure before you pick the detector.

1. Route dialogue tracks to a dialogue bus and music tracks to a music bus with the Fairlight bus assignments.
2. Open Dynamics on the music bus.
3. Enable the Compressor.
4. In the compressor's Sidechain area, switch the sidechain On.
5. Pick the dialogue bus as the sidechain source.
6. Use Listen briefly to confirm the detector hears the right dialogue bus, then switch Listen off again.
7. Set Threshold and Ratio for the amount of gain reduction, then shape the move with Attack, Hold and Release.

Threshold decides when the detector starts the compression. Ratio sets how hard the music gets pulled down once the signal crosses that threshold. Attack controls the entry, Hold keeps the reduction steady, and Release controls the return. The Fairlight compressor guide walks through the same controls in a dialogue context. Watch the gain-reduction meter, but make the call with your ears.

A sidechain isn't an audible cable from the dialogue into the music. The dialogue bus stays in the normal mix; a detector feed measures it and drives the compressor on the music bus. Leave Listen on by accident and you may hear only the detector and decide the routing is broken.

Where AI Audio Assistant Fits

AI Audio Assistant is a whole-mix starting point in Resolve Studio, not just a newer Ducker button. The current Resolve 21 Reference Manual says it can organize and color-code tracks, even out dialogue levels, set the mixer faders, clean up dialogue, apply ducking to music tracks, write automation and add the finishing plugins for the delivery standard you choose.

Only use it once the audio elements sit on separate tracks with the right track categories. Run Timeline > AI Tools > Audio Assistant on an unmixed duplicate timeline, pick the Delivery Standard and click Auto Mix. Then play the whole result, inspect the automation it wrote and compare it with the untouched version.

If a track landed in the wrong category, right-click its track header, fix it under Track Category and rerun the tool. If all you need is one music bed under one voiceover, Ducker is faster and much easier to read.

Don't run Audio Assistant, Ducker and a bus sidechain blindly on the same music path. Their reductions can add up until the bed collapses. Bypass them one at a time and keep the system whose result is easiest to control.

Use Keyframes for Editorial Exceptions

Automatic detection can't know that a musical hit should stay loud under a quiet line, or that the music needs to rise before the speaker finishes. For those moments, draw clip volume keyframes or record fader automation. On the Edit page, Option-click the music clip's Volume overlay line (Alt-click on Windows) to place four keyframes, two for the entrance and two for the return, then drag the segment between the middle pair down. To soften those bends, select the keyframes, right-click one and choose Ease In, Ease Out or Ease In and Out.

A practical hybrid looks like this:

1. Let Ducker or sidechain compression handle the consistent baseline.
2. Bypass or soften it for a shot where the music has to lead.
3. Shape that exception with keyframes that start before the words and finish after them.
4. Check the handoff into the next automatic section for a level jump.

Manual moves take longer, but they follow the story instead of a threshold. Use them where meaning matters more than repeatability.

Fix Pumping, False Triggers and Bad Renders

The music pumps between words

Raise Hold or lengthen Recovery Time or Release so short gaps in the speech don't bring the full bed back. If the reduction is deep, make Duck Amount milder or lower the compressor ratio.

The first syllable is masked

Add enough Lookahead in Ducker, or shorten the compressor's Attack. Also check that the trigger track starts with the actual consonant and doesn't have a late clip fade.

Breaths and room noise trigger the music

Clean or edit the source, then check again. On a compressor sidechain, set Threshold so real speech crosses it and the noise floor doesn't. Don't gate so hard that you clip the ends of words.

The music never returns

Solo or Listen to the detector source. A noisy dialogue track can stay above threshold the whole time. Check Hold and Recovery Time or Release too, and make sure no other dialogue source is continuously active.

The timeline is clean but the render crackles

Render a short problem range, then bypass Ducker or sidechain processing and render the same range again. That A/B test tells you whether the processor path is involved. It doesn't prove a general Resolve bug.

If the fault follows the processor, save a duplicate project and rebuild that delivery move with keyframes or recorded automation. Check both files in an external player before you touch the rest of the mix.

Finish the Mix After Ducking

Ducking creates separation; it doesn't set final program loudness. Listen on headphones and small speakers, read the loudness meters and check the full mix for peaks after all busses and processing. The Resolve LUFS and audio-level guide handles that delivery pass.

Before the final render, export a short section that represents the whole piece: a strong music entrance, normal dialogue, a pause and the loudest transition. If that section is clean, render the full program and listen to the result outside Resolve. The audio export guide covers stems and audio-only files.

Frequently Asked Questions

Is audio ducking available in DaVinci Resolve Free?

Yes. Ducker Track FX is in the free edition; AI Audio Assistant is Studio-only. For a controlled mix without AI, Fairlight's regular mixer and Dynamics tools give you the non-AI route.

Which track gets the Ducker effect?

Put Ducker on the track you want to lower, usually the music, and choose the dialogue as its source. The source triggers the change; the target receives it.

What is the difference between Ducker and sidechain compression?

Ducker is a purpose-built Track FX with direct depth and timing controls. Sidechain compression runs through the Fairlight Dynamics compressor and can use a track or a bus as its detector, which makes it the better fit for bus-based mixes.

Can several voice tracks trigger one music Ducker?

Yes. Current Resolve versions let you add several source tracks to one Ducker, which helps when dialogue is split across speakers, microphones or voiceover tracks.

How much should music duck under dialogue?

There's no universal value. The manual's default is 2.7 dB, and it suggests 2 to 5 dB for most cases. Duck only until the words stay clear on the devices people will actually use. Deep, fast moves sound like pumping, and dense music may need more room than a sparse bed.

Should I use AI Audio Assistant instead?

Use AI Audio Assistant in Studio when you want a first-pass mix across categorized dialogue, music and effects tracks. Use Ducker or sidechain compression when you want predictable control over one specific relationship.

The Fairlight audio hub covers the whole path, from track organization and dialogue cleanup through loudness, bus routing and delivery.

Jason Miller I run DaVinci Resolve Club, an independent publication. My guides explain editing, color, Fairlight, Fusion and delivery, with clear Free vs Studio limits.