Vizard Agent

How to drop an ad into a podcast at a natural pause

Last updated 2026-09-05 · 8 min read

Give Vizard Agent the episode and the ad and say roughly where it should go. It looks for a point that is both a change of subject and a genuine pause, matches the ad's loudness to your speaking voice, trims the silence off its ends, joins it in and then has both transitions listened to rather than assumed.

What is the short version?

An ad dropped at a timestamp lands mid-thought. An ad dropped at a pause lands mid-sentence surprisingly often, because pauses happen inside sentences too. What you want is a pause that is also a change of subject, and that is found in the transcript rather than in the waveform.

  1. Go to Vizard Agent with the episode and the ad.
  2. Say roughly where it should sit.
  3. Say it should land at a natural break.

What do you need before you start?

Two files and a rough window. You do not need to find the spot — that is the work — but saying "somewhere in the first forty-five minutes" narrows the search to the part of the episode where an ad is worth anything.

What do you type into Vizard Agent?

Ask Vizard Agent for a natural break rather than for a time. A time is easy to satisfy and usually wrong; a natural break is a condition the transcript can actually be searched against, and it is what you meant in the first place.

Prompt

Here is an ad — can you add it somewhere there is a natural pause within the first 45 minutes of the episode? Match its level to my voice so it does not jump, and give me an mp3 as well as an mp4 when you are done.

Variants worth knowing:

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real podcast episode that needed an ad inserted after it had already been cleaned up. The ad itself is inspected before any decision is made about where to put it.

  1. Looks at the ad file and views frames from it.
  2. Finds section changes with a real pause inside your window.
  3. Measures the loudness of your speaking voice and of the ad.
  4. Checks the ad's start and end for dead air to trim.
  5. Renders the episode halves and the conformed ad together.
  6. Joins the pieces into the finished episode.
  7. Cuts out the transition region and uploads it for review.
  8. Has both transitions listened to rather than assumed.
  9. Delivers the finished episode with the ad in place.
  10. Exports the mp4 and the mp3 from the same master.

Step two is the whole decision. A pause on its own is a bad target because people pause mid-sentence constantly; a pause that coincides with a change of subject is a place where a listener is already between two things, and an ad there costs almost nothing.

Step five is quietly efficient. The two halves of the episode and the level-matched ad are rendered at the same time rather than one after another, so a long episode does not have to be encoded twice to get one break into it.

Step four is the small one that saves an awkward join. Ads arrive with silence on the front and back from whoever exported them, and that silence lands as a hole either side of the break unless somebody goes looking for it.

What does the result look like?

An episode with the ad sitting at a moment where the conversation had just finished a thought, at the same loudness Vizard Agent measured on your voice, with no gaps around it and no clipped word on either side of the join.

You get the video and the audio file from the same master, so they cannot disagree.

When does this not work well?

Some episodes do not have a good break. A continuous interview, a tightly edited narrative, or a conversation that never changes subject will force the ad somewhere that costs more, and it is better to know that than to have one inserted anyway.

How do you fix a result that came back wrong?

Say what you hear at the break. Vizard Agent keeps the transcript, the loudness measurements and the join points, so moving the ad or fixing a level is a change to the assembly rather than a re-render of the whole episode from scratch.

How does Vizard Agent compare to doing it yourself?

By hand you scrub to about the right place, find a gap in the waveform and drop the ad there. That gap is frequently a breath in the middle of a sentence, and the ad's own padding then adds a second of silence on top of it.

By hand Vizard Agent
Where it lands A gap in the waveform A pause at a section change
The ad's padding Left on Trimmed from both ends
Levels Eyeballed Measured against your voice
Checking the join Played once Both transitions listened to
Deliverables Exported separately mp4 and mp3 from one master

Common questions

Can it find the spot itself? Yes. That is the point — Vizard Agent searches the transcript.

Can I give a fixed timestamp instead? Yes. Vizard Agent will place it exactly where you say.

What about several ads? Say how many. Vizard Agent spaces them at section changes.

Will the ad be louder than me? Not once matched. Vizard Agent measures both.

Does it trim the ad? Only the silence at its ends, unless you ask Vizard Agent for more.

Can I hear the join before delivery? Yes. Vizard Agent exports the transition region for you.

Do I get an audio-only version? Yes. Ask Vizard Agent for the mp3 alongside the mp4.

Can it add a jingle around the ad? Yes. Supply it and Vizard Agent places it either side.

What if my episode has music under it? Vizard Agent will say when a clean join is not possible.

Will it tell me where it put it? Yes. Ask Vizard Agent for the timestamp and the reason.

Can it do this on a long back catalogue? Yes. Vizard Agent applies the same rule episode by episode.

What about a mid-roll and a pre-roll? Say which is which and Vizard Agent places both.

Does the episode get re-encoded? Only around the joins where it must.

Why not just use the longest pause? Because the longest pause is often somebody thinking mid-sentence, and an ad there cuts them off.