How to make an added voice sound like it is really in the scene

Last updated 2026-09-05 · 8 min read

Tell Vizard Agent the added voice sounds like it was recorded next to your ear. It rebuilds that line at a greater distance rather than simply turning it down, mixes it inside the location's existing sound instead of laying it on top, and confirms on the waveform that the running water or the market noise still carries underneath.

What is the short version?

"Too close to the microphone" and "too loud" are different complaints, and only one of them is fixed by the volume control. A quiet close voice still sounds like it is in your headphones; a distant voice sounds like it is in the shot.

Go to Vizard Agent with the clip.
Say the voice sounds too close, not too loud.
Say what location sound has to survive underneath.

What do you need before you start?

The footage with its own sound intact, and a description of where the speaker is meant to be. "Further away", "across the courtyard", "somewhere behind the noise" all describe a position, and a position is something a mix can actually be built towards.

The clip with its original audio. Not a stripped version.
Where the voice is coming from. Near, far, off screen.
What must stay audible. Water, traffic, a crowd.
Who is speaking. Adult, child, a group answering.
Which language or dialect. It changes the casting.

What do you type into Vizard Agent?

Say the word "distance" in the message itself. Asking Vizard Agent for a quieter voice gets you a quieter voice; asking for a more distant one gets you a different take, mixed a different way, and that is the thing that actually solves the problem you are hearing.

Prompt

The greeting and the reply sound like they were recorded right on the microphone — they should be further away, as if the speakers are actually out there in the shot. Keep the sound of the water running underneath the whole time; the speech should sit inside it, not on top of it.

Variants worth knowing:

"Further away, not quieter." Names the real problem.
"Inside the noise, not on top." Describes the mix you want.
"Keep the location sound." Protects what makes it real.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real outdoor clip that needed spoken lines added over running water. Notice how much measuring happens before anything is placed, and that the location sound is treated as a thing to protect rather than a thing to duck.

Examines the opening to find where the lines should sit.
Listens to the existing audio to hear what has to survive.
Casts voices suited to the speakers and the dialect asked for.
Generates the lines and measures their length and level.
Listens back to them to confirm they are clear enough to place.
Composites them onto the video while preserving the location sound.
Verifies the balance numerically between the speech and the water.
Reviews the waveform to confirm the ambience runs unbroken.
Rebuilds from the original with a more distant voice when asked.
Verifies the dialect, the timing and the distance on the render.
Generates a separate sound effect and checks it is what was asked for.
Replaces it with a real recording when the generated one is not.
Confirms on the waveform that the effect lands where it should.

Step nine is the whole article in one line. The fix for a voice that is too close is a new take built to sit at a distance, composited back from the original file, rather than the same take with the fader pulled down.

Steps eleven and twelve are the trap next door. A short generated effect can come back as a piece of music that gestures at the sound you asked for, so the effect gets listened to on its own before it goes anywhere near the mix, and replaced with a real recording when it is not the thing.

What does the result look like?

Speech that belongs to the place: it arrives from somewhere in the scene rather than from the front of the mix, the water or the market keeps running underneath it without dipping, and the whole thing sounds recorded rather than assembled.

The give-away, when it is wrong, is that everything is equally close. When it is right, nothing is.

When does this not work well?

Distance is built from reflections and from what the location sound does around a voice, so a scene with no ambience of its own gives very little to work with. Clarity also fights realism, and sometimes clarity has to win.

Silent locations. Nothing for the voice to sit inside.
Dialogue that must be understood. Distance costs intelligibility.
Very short lines. No time for the space to register.
Loud continuous noise. A distant voice will be lost in it.
Existing close-mic dialogue. Mixing near and far reads oddly.

How do you fix a result that came back wrong?

Say whether the problem is position or level, because those get fixed differently. Vizard Agent keeps the takes, the measured levels and the ambience it preserved, so a correction changes one of those rather than rebuilding the whole soundtrack.

"Still too close." Rebuilt further back, not turned down.
"Now I cannot hear it." Brought forward while keeping the space.
"The water disappeared." Ambience restored under the line.
"That effect is music." Replaced with a real recording.

How does Vizard Agent compare to doing it yourself?

The manual fix is the fader, which changes loudness and not position, so the voice stays in the front of the mix and simply gets harder to hear. The location sound then tends to get ducked to make room, which removes the very thing that made the shot sound real.

	By hand	Vizard Agent
"Too close"	Turned down	Rebuilt at a distance
The location sound	Ducked under the voice	Preserved and measured
Where the voice sits	On top of the mix	Inside the ambience
Checking it	By ear	Levels measured, waveform read
Sound effects	Used as generated	Listened to, replaced if wrong

Common questions

Is this different from turning it down? Yes, entirely. Vizard Agent rebuilds the take rather than the level.

Will the location sound survive? Yes. Vizard Agent measures it and mixes the voice inside it.

Can it do a group of voices? Yes. Say how many and how far, and Vizard Agent casts them.

What about a specific dialect? Say which. Vizard Agent casts towards it and you can check.

Can I supply a reference voice? Yes. A sample is the clearest instruction Vizard Agent can get.

Will it duck the ambience? Not unless you ask. Ducking is what makes it sound added.

Can it place a voice off screen? Yes. Say where the speaker is and Vizard Agent mixes for it.

What if the effect sounds wrong? Say so. Vizard Agent will go and find a real recording instead.

How do I know the ambience is still there? Vizard Agent shows you the waveform with it running underneath.

Can it match an existing voice in the clip? Yes, if Vizard Agent has a clean sample to work from.

Does distance work for music too? Yes. Ask Vizard Agent for a band heard from another room and it is the same idea.

Will this survive on phone speakers? Partly. Say if phone playback matters and Vizard Agent leans closer.

Can it make a voice sound indoors? Yes. Name the space and Vizard Agent builds towards it.

Why not just record it quieter? Because quiet and distant are different sounds, and only one of them puts the speaker in your shot.