Give voice direction without rewriting approved narration

A direction such as “quietly, with relief” is not part of the sentence an audience should hear. Store it beside the sentence, not inside text that might be read literally by a speech engine.

VocalLab presents controllable voice delivery as part of its voice studio. That is a useful interaction reference, but a marketing example does not establish that every provider accepts the same instruction syntax or produces the same result. Official product overview.

Use three separate records

Keep exact spoken text, pronunciation guidance and performance direction separate. The first record determines factual wording. The second explains how names or unfamiliar terms should sound. The third describes pacing or emotion. A change to any of them can require a new audio version, even if the displayed script looks similar.

Avoid ambiguous notes such as “make this stronger.” Say what a listener should notice: a longer pause before the result, restrained surprise at the discovery or a clear contrast between two numbers. These are review instructions rather than a guarantee that a model will follow them.

A worked example

The fictional approved line is “The lamp is still dark.” A pronunciation note is unnecessary. A delivery note says “Matter-of-fact, without alarm; pause after dark.” The next line, “Now the circuit is complete,” should sound like a result, not a dramatic revelation that the story has not earned.

Do not insert stage directions in parentheses into the approved narration and assume they will be ignored. Some systems read them, others interpret them and others reject them. Check the selected provider's documented input format before sending any instructions.

Review a short representative passage

Choose a passage containing the hardest name, a number and an ordinary sentence. Listen for pronunciation, pacing and level consistency. A voice sample proves that an audition entry plays; it does not prove that your complete script will be pronounced correctly.

Use the voice over script template to keep your own spoken sections clear. Its duration estimate uses a chosen rate and does not produce audio. The prosody glossary explains why rhythm and emphasis can change perceived meaning even when every word is retained.

Budget the words before adding direction

For a 60-second edit, reserve 10 seconds for pauses and moments when the picture can speak for itself. At 140 words per minute, the remaining 50 seconds fit about 117 words. One starting split is 15 for the hook, 65 for the explanation, 12 for the transition and 25 for the closing. Replace those targets with the times from a read aloud. Write notes such as “hold on the circuit for two seconds” next to the line so they stay out of the spoken word count.

After the read, put the exact words and visual directions in separate columns of an AV script. The visual column tells the editor what to show during each pause.

Review before handoff

Record which text and direction version produced the accepted audio. When one line changes, identify dependent captions, timing and shots before replacing it. A shorter take can create a visible hold; a longer take can push the next shot out of alignment.

Keep the earlier audio available for comparison until the replacement is checked in context. Local correction is useful only when it preserves the parts already accepted and makes any newly invalid timing obvious. A confident-sounding voice is never a substitute for checking the underlying claim.

References checked November 4, 2026. Product descriptions are from public official material, not paid-output testing. Examples and cover artwork are original Topicut editorial material.