Vizard Agent

How to give a generated presenter the sound of the room they are standing in

Last updated 2026-09-10 · 8 min read

Tell Vizard Agent which room the presenter is supposed to be in and ask for room tone under the voice. It generates a matching ambience, measures the narration and the ambience against each other before mixing, and sits the tone far enough down that you never consciously hear it — you only stop hearing the silence that gave the shot away.

What is the short version?

A generated presenter is filmed in nothing. There is no air in the room, no fridge, no traffic, no floor — just a clean voice against pure digital silence, and viewers read that absence as fake long before they can say why.

  1. Go to Vizard Agent with the ad.
  2. Name the room the presenter is meant to be in.
  3. Ask for room tone under the voice, measured.

What do you need before you start?

Not much beyond a decision about where this person is supposed to be standing. The room is the only creative input that matters here, and Vizard Agent needs it named rather than implied, because a kitchen, an open-plan office and a bedroom sound nothing alike.

What do you type into Vizard Agent?

Say it as a note about the room, not as a request for a sound effect. A sound effect is an event you notice; room tone is a floor you do not, and Vizard Agent mixes the two very differently.

Prompt

Make the UGC ad as described. The presenter is in a normal flat, so put subtle room tone under the whole thing — I want the voice to sound like it was recorded in that room, not in a booth. Measure it against the narration so it never draws attention.

Variants worth knowing:

What does Vizard Agent actually do?

Here is the shape of it on a real vertical ad with a generated presenter. The important thing is that the tone is treated as a mix decision with a measured level, not as one more asset dropped onto the timeline at whatever volume it arrived at.

  1. Reads the room you named from the brief.
  2. Generates a subtle indoor ambience to match it.
  3. Checks it is a bed, not an identifiable noise.
  4. Measures the narration's loudness.
  5. Measures the ambience against it.
  6. Sets the tone well under the voice.
  7. Extends it across the product inserts so it does not drop out.
  8. Carries it through the cuts rather than restarting per clip.
  9. Renders a voice-led mix with the tone underneath.
  10. Checks the waveform to confirm the bed survived the render.

Step five is the one people skip. Ambience generated at its own natural level is far too loud for a bed, and Vizard Agent measuring both sides before mixing is what turns a hiss you notice into a room you do not.

Step seven matters more than it sounds. The tone dropping out under a product insert is worse than never having it, because the silence arrives suddenly and the viewer hears the edit rather than the room. Vizard Agent runs the bed under the inserts as well.

Step eight is the same problem across cuts. A tone that restarts at every clip boundary ticks audibly, so Vizard Agent lays one continuous bed under the assembled cut instead of one per shot.

What does the result look like?

An ad that sounds like it was shot somewhere. Played back to back with the untreated version the difference is hard to name — most people say the second one sounds "more real" without identifying a single sound.

Turn the tone off and the fakeness comes straight back, which is the useful test.

When does this not work well?

Room tone fixes one specific thing, which is the absence of a room. It does not fix a voice that is wrong for the space, it will not rescue a mix that has other problems underneath it, and there are cuts short enough that Vizard Agent adding a bed makes no perceptible difference at all.

How do you fix a result that came back wrong?

Describe what you can actually hear, in those words. If you can hear the room tone at all as a distinct sound, then the level is wrong, and that is by far the most common note on a first pass — Vizard Agent will bring it down rather than replace it.

How does Vizard Agent compare to doing it yourself?

By hand this is a job people know about and skip, because finding a usable tone, trimming it to length and finding its level is twenty minutes of work for something no viewer will ever compliment. Vizard Agent does it as part of the mix rather than as an optional extra.

By hand Vizard Agent
Finding a tone Search a library Generated to the named room
Setting the level By ear, hopefully Measured against the narration
Under inserts Often forgotten Carried through
Across cuts Restarts and ticks One continuous bed
Verifying Trust the render Waveform checked

Common questions

Will I hear it? You should not. Vizard Agent sets it to be missed, not noticed.

Can I ask for a specific room? Yes. Name it and Vizard Agent generates to match.

Does it work with an avatar presenter? Yes. That is exactly the case it was built for.

What if I also have music? Say so. Vizard Agent balances the two rather than stacking them.

Can it go under the inserts too? Yes, and by default Vizard Agent carries it through.

Will it tick at the cuts? Not if Vizard Agent lays one bed under the whole assembly.

What level should it be? Well under the voice. Vizard Agent measures rather than guesses.

Can I hear the version without it? Ask Vizard Agent for both. The comparison is the point.

Does it help with captions? No, they are unrelated. This is purely a mix question.

What about outdoor scenes? Same idea. Tell Vizard Agent the street, not the room.

Can it use my own recording? Yes, if you have real tone from the location, give it to Vizard Agent.

Does it change the voice? No. The voice is untouched; only what sits under it changes.

Is this the same as ambience effects? No. Effects are events. Vizard Agent treats tone as a floor.

Why does silence read as fake? Because no real room is silent, and the ear knows that even when the viewer does not.