How to give a narration room to breathe between its sections

Last updated 2026-09-10 · 8 min read

What is the short version?

A generated narration comes back as one unbroken paragraph of speech. Every section runs into the next at the same pace, so the viewer never gets the half-second that tells them one idea has finished and another has started.

1. Go to Vizard Agent with the script.
2. Ask for gaps between the sections, not one run.
3. Ask for room at the end for the last line.

What do you need before you start?

The script and its structure. Vizard Agent puts the silence where the argument turns, so it needs to know where the turns are — a script written as one block gets narrated as one block, however good the voice is.

• The script, with its sections marked. Headings are enough.
• The target length. Gaps cost seconds.
• Where the picture changes. They should agree.
• Whether music runs underneath. It fills the gaps.
• How the video ends. A fade needs more room than a cut.

What do you type into Vizard Agent?

Say it feels breathless rather than fast. Those are different complaints with different fixes — slowing the voice down makes a rushed narration sound laboured, whereas putting gaps between the sections leaves the pace alone and gives the structure back.

Prompt

Variants worth knowing:

• "Breathless, not fast." Points at structure.
• "Real gaps between sections." Silence, not slower speech.
• "Room at the end." The most common miss.

What does Vizard Agent actually do?

Here is how it goes on a real explainer built entirely from generated narration. The important thing to notice is that the gaps are part of how Vizard Agent assembles the track in the first place, rather than something spliced into a finished recording afterwards.

1. Reads the script's structure, section by section.
2. Generates the narration for each section.
3. Assembles the track with breathing room between them.
4. Measures the speech pauses around each planned cut.
5. Aligns the picture changes to those pauses.
6. Checks nothing lands mid-sentence.
7. Extends the ending so the last line is not clipped.
8. Renders and reviews the waveform for the gaps.
9. Confirms the music does not fill every one of them.

Step three is the difference between a fix and a design. Cutting silence into a finished narration track leaves audible seams, whereas building the sections separately and spacing them means each gap is genuine room rather than a splice.

Step seven is the one that catches almost everybody. A video that ends the instant the last syllable finishes reads as truncated, and the repair is a second or two of picture that the edit never planned for.

Step four is what stops the gaps being wasted. A pause in the speech is only useful if the picture uses it, so Vizard Agent measures the real speech pauses around every planned cut and moves the cut into one rather than landing it across a word.

Step nine is worth saying out loud. A music bed running at full level through the pauses fills the space you just made, so Vizard Agent checks that the gaps are actually audible rather than only present in the speech track.

What does the result look like?

The same script, the same voice and very nearly the same length, that no longer feels like it is racing to get to the end. The sections read as sections, the picture changes sit in the pauses, and the last line has somewhere to land.

Vizard Agent shows you the waveform of the result, where the gaps are visible as gaps rather than as one continuous block of speech.

When does this not work well?

Not everything wants air. A fast-cut social edit is breathless on purpose, and a hard duration limit may not have two spare seconds anywhere in it, in which case something has to come out of the script instead.

• Hard length limits. Gaps cost time you may not have.
• Deliberately relentless edits. Breathlessness is the style.
• Wall-to-wall music. The gaps will not be heard.
• Very short videos. Under twenty seconds, there is no room.
• Dense scripts. Cut words first, then add air.

How do you fix a result that came back wrong?

Point at the place where it still runs on. Because Vizard Agent built the track section by section rather than as one recording, adding air at one particular turn stays a local change instead of forcing a re-timing of the whole narration.

• "Still rushed at the second turn." A longer gap there.
• "Now it drags." Vizard Agent tightens the gaps.
• "The music covers the pauses." The bed ducks.
• "The ending still feels cut off." More room after the line.

How does Vizard Agent compare to doing it yourself?

By hand the usual response to a breathless narration is to slow the voice down, which makes it sound laboured without giving any of the structure back. The ending tends to get fixed last of all, after somebody watching it says it feels abrupt.

Common questions

How long should a gap be? Usually under a second. Vizard Agent sizes it against the pace.

Will it change the voice? No. Vizard Agent leaves the delivery alone and adds space around it.

Does it make the video longer? A few seconds. Tell Vizard Agent if you have a hard limit.

Can it take words out instead? Yes, and on a tight limit Vizard Agent will suggest that first.

Does the music need to duck? Often. Vizard Agent checks whether the bed is filling the gaps.

What about the ending specifically? Vizard Agent extends it so the last line is not clipped by a fade.

Will the picture changes still line up? Yes, Vizard Agent aligns them to the pauses it created.

Can I mark the sections myself? Yes, and that gives Vizard Agent the best result.

Is this the same as slowing it down? No. The speech runs at its natural pace with room around it.

Does it work on a recorded voice? Yes, though the gaps are cut in rather than built.

How do I see the gaps? Ask Vizard Agent for the waveform; they are visible in it.

What if it now drags? Say so and Vizard Agent tightens them.

Does it apply to captions? They follow the speech, so they get the same rhythm.

Why does it feel breathless at all? Because generated speech has no reason to pause unless it is asked to.