How to duck the original audio under sound effects you add

Last updated 2026-09-05 · 8 min read

What is the short version?

Adding a sound on top of a clip that already has sound gives you two things competing. The fix is not a louder effect, it is a decision about what the original does at that moment — quieter, or gone — and those are two different instructions.

1. Go to Vizard Agent with the clip and the sounds.
2. Say where each one lands.
3. Say whether the original dips or disappears there.

What do you need before you start?

The clip, the sounds and a moment for each. Moments described in the action — "when he starts to move", "when the line about the score comes in" — work better than timestamps, because they survive a recut and a timestamp does not.

• The clip with its own audio.
• The sounds. Files or descriptions.
• Where each goes. By what happens on screen.
• What the original does. Dip or mute.
• What it comes back to. "Normal, not loud."

What do you type into Vizard Agent?

Give the behaviour, not just the placement. "Put this sound here" leaves the original audio's fate unstated, and whatever gets chosen will be wrong half the time — saying "duck the original, then bring it back" removes the guess entirely.

Prompt

Variants worth knowing:

• "Lower it, then bring it back." A dip with a recovery.
• "Take it out completely." A mute, not a dip.
• "Back to normal, not too loud." Sets the return level.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real gameplay edit that needed several sound effects placed over its own audio. The mix gets rebuilt twice, and the second rebuild is a different instruction rather than a better attempt at the first one.

1. Transcribes the supplied sound clips to know what each says.
2. Inspects the effect files and previews them.
3. Measures the level of the source audio and of every effect.
4. Places the effects and renders a first mix.
5. Reviews the finished waveform for gaps and spikes.
6. Measures the source and the effects again before rebalancing.
7. Re-balances with a temporary dip in the original audio.
8. Prevents the dip overshooting so nothing jumps on the way back.
9. Checks the balance and the timing of each added line.
10. Mutes the original entirely during the added lines instead.
11. Confirms it disappears during them and returns after.
12. Checks frames and waveform with the two tracks separated.

Step three is why the mix holds together. An effect placed without measuring is either buried or startling, and the level it should sit at depends on what the clip's own audio is doing at that exact moment rather than on a general setting.

Steps seven to eleven are two different answers to the same request, tried in order. A dip keeps the clip feeling continuous and works for most effects; a full mute is what you need when the added audio is a line of speech, because a dipped original still competes for the same attention.

What does the result look like?

Each sound landing exactly where you asked for it, clearly audible, with the clip's own audio stepping back underneath and returning afterwards without a jump. Where you asked Vizard Agent for silence, the original is genuinely gone rather than merely quiet.

Nothing sounds pasted on, which is the whole test.

When does this not work well?

Ducking assumes there is something to duck. Continuous loud audio — a crowd, an engine, music with no gaps — leaves the effect fighting for room, and a dip deep enough to make space becomes audible as a hole.

• Loud continuous source audio. The dip becomes obvious.
• Effects that overlap speech. Mute is usually the answer.
• Very short effects. There is no time to dip and return.
• Music you must not interrupt. Then the effect has to fit around it.
• Many effects close together. The original never comes back up.

How do you fix a result that came back wrong?

Say whether you want less of the original or none of it. Vizard Agent keeps the measurements and both mixes, so changing a dip into a mute at one moment is a change to that moment rather than to the whole soundtrack.

• "I still hear the original under it." Muted there instead.
• "It jumps when it comes back." The recovery smoothed.
• "The effect is too loud." Re-measured against the source.
• "Bring the original back sooner." The dip shortened.

How does Vizard Agent compare to doing it yourself?

By hand, effects get dropped in at a level that sounded right on the first listen and the original is either left alone or turned down for the whole clip. Both produce the same complaint, which is that the added sounds feel stuck on top rather than part of it.

Common questions

Can it place sounds by what happens on screen? Yes. Describe the moment and Vizard Agent finds it.

Do I have to supply the sounds? No. Vizard Agent can generate them if you describe them.

What is the difference between dip and mute? A dip leaves the original quietly present; a mute has Vizard Agent remove it.

Can it do both in one video? Yes, and Vizard Agent finds that is usually what you want.

Will the dip be audible? Not if it is shaped properly. Vizard Agent smooths the recovery.

Can I set how far it dips? Yes. Say how much quieter and Vizard Agent works to it.

What about music underneath? Same treatment. Tell Vizard Agent which layer should duck.

Can it stop the level jumping back up? Yes. Vizard Agent limits the recovery so nothing overshoots.

Will it tell me the levels it used? Yes. Ask Vizard Agent for the measurements.

Can it match a meme edit's style? Yes, if you point Vizard Agent at an example.

What if my effect is a whole spoken line? Then mute the original under it; Vizard Agent will suggest that.

Does the video change? No. Vizard Agent changes the audio only.

Can it export the effects separately? Yes. Vizard Agent exports them as their own track for remixing later.

Why measure rather than set a level? Because the right level for an effect depends entirely on what the clip is already doing when it lands.