How to write a script that fits the runtime before you record it

Last updated 2026-09-17 · 8 min read

What is the short version?

A script written to a target length is written to an estimate of how fast someone will read it. Voices differ by a lot, and the error only shows up once the recording exists and the film is eight seconds long.

1. Ask Vizard Agent to choose the narrator first.
2. Ask it to measure that voice's real speaking pace.
3. Ask it to tighten the script to the measured number.

What do you need before you start?

The script in draft and a firm runtime to hit. Vizard Agent measures the writing against a target, so a soft brief like "about a minute" can only produce a correspondingly soft result — the whole discipline depends on the number being a real constraint.

• A draft script. It does not have to be final.
• The runtime. Exact, or a narrow range.
• The voice. Or let it audition some.
• What must survive. The lines you will not cut.
• Any pauses you want. They count against the total.

What do you type into Vizard Agent?

Tell Vizard Agent the order of operations you want. Left to itself the natural order is script first, then voice, then discovering the problem at the end; asking for the voice to be cast first is the thing that makes the measurement possible at all.

Prompt

Variants worth knowing:

• "Cast the narrator first." Reverses the usual order.
• "Measure how fast that voice actually reads." Not an average.
• "Keep the closing line." Says where the cuts come from.

What does Vizard Agent actually do?

Here is the Vizard Agent sequence on a real explainer that had to land on a fixed runtime. The two measurement steps both sit before the recording rather than after it, and that ordering is the whole difference between this and the usual way round.

1. Prepares the narration text and counts the words.
2. Casts a narrator voice for the piece.
3. Measures that narrator's real speaking pace.
4. Searches for the footage each section will need.
5. Tightens the script to fit the runtime at the measured pace.
6. Fills the remaining visual gaps now the sections are sized.
7. Finalises the tightened narration.
8. Generates every narration segment.
9. Takes word-level timings from the recorded audio.
10. Builds the cards and cut against those real timings.

Step three is the step that is normally skipped. Two voices reading the same paragraph can differ by fifteen per cent, so a script sized on a general figure is sized for a narrator you did not cast.

Step five is cheaper than every alternative. Cutting forty words from a draft takes a minute; discovering the overrun after recording means either re-recording, speeding the voice up until it sounds hurried, or losing the end.

Step nine is a second measurement, not a repeat of the third. The pace estimate sizes the script; the word-level timings from the actual recording are what the pictures get cut to.

What does the result look like?

A film that lands on its runtime with the ending intact. The narration was written to fit rather than trimmed to fit, and Vizard Agent tells you the measured pace and the final word count so the next script starts from a real number.

When does this not work well?

Pace is not constant even within one voice. An emphatic passage runs slower than a list, so a single figure across a long script drifts, and for anything past a couple of minutes the measurement has to be taken per section.

• Long scripts. One pace figure drifts over several minutes.
• Mixed delivery. Emphasis and lists run at different speeds.
• Unscripted presenters. There is nothing to measure against.
• Soft targets. Without a firm runtime there is nothing to fit.
• Heavy pauses. They have to be budgeted separately.

How do you fix a result that came back wrong?

Say whether it overran or underran, and by how many seconds. Vizard Agent still holds the measured pace for that voice, so it can convert those seconds straight into a word count and tell you exactly how much writing has to come out.

• "It is eight seconds long." Roughly twenty words, cut from where you say.
• "It finishes early." The pauses widen or a line is restored.
• "The read sounds rushed." The pace was pushed; the script is cut instead.
• "The ending got trimmed." The cuts are moved to the middle.

How does Vizard Agent compare to doing it yourself?

By hand you write to a words-per-minute figure from an article, record it, and find you are over. Then the options are all bad: speed the read up, cut the conclusion, or let the film run long — and the one people pick is speeding up the read.

Common questions

How wrong is a standard figure? Enough to cost several seconds a minute. Vizard Agent measures rather than assuming.

Can it audition voices first? Yes, and their paces differ. Vizard Agent reports each one.

What if I cannot cut anything? Then the runtime has to move. Vizard Agent will say by how much.

Where should the cuts come from? Wherever you name. Tell Vizard Agent which lines are untouchable and it cuts elsewhere.

Can it slow a voice down instead? A little, before it sounds odd. Vizard Agent will tell you where that line is.

Does this work for a presenter on camera? Only if they follow the script. An unscripted read cannot be measured in advance.

What about pauses? They count. Tell Vizard Agent what you want and it budgets for them.

Can it measure per section? Yes, and for long scripts Vizard Agent should. A single figure drifts.

Will it rewrite my words? Only to tighten, and only if you ask. Otherwise it tells you the overrun.

What is the second measurement for? Word-level timings from the real recording, which the pictures are cut to.

Can I reuse the pace next time? Yes, for the same voice. Vizard Agent reports it so you can.

Does music affect the runtime? It affects the feel, not the length. The narration sets the length.

What if the script is already recorded? Then it is a re-edit rather than a rewrite. Vizard Agent works from the timings.

How accurate does the runtime need to be? As accurate as your delivery slot. Vizard Agent treats the number as given.