Voice-over tone and delivery direction for a script

You are a voice director preparing a script for recording. Add delivery direction to the script below.

Who will read it: [e.g. "a human narrator", "a text-to-speech tool", "me, recording at home"]
Project type and audience: [e.g. "60-second product explainer for small business owners"]
Desired tone: [e.g. "warm and confident, like a helpful colleague, not a salesperson"]
Target length: [e.g. "55-60 seconds"]
Words that need pronunciation guidance: [NAMES, ACRONYMS, FOREIGN TERMS]
Markup supported by the tool (if TTS): [e.g. "SSML", "plain text with punctuation only", "unknown"]

Script:
"""
[PASTE SCRIPT]
"""

Produce:
1. A one-paragraph direction for the whole read: character, energy, pace, what to avoid.
2. The estimated duration at the suggested pace (show words per minute) and lines to trim if it runs long.
3. The script split into short segments. For each: tone note, pace (slower/normal/faster), words to stress in CAPITALS, pauses marked as / (short) and // (long).
4. A pronunciation guide using simple respelling (e.g. "NOO-ruhl") and notes on acronyms (spell out or say as a word).
5. If a TTS tool is used: a version formatted for the markup it supports, or punctuation-only tips if markup is unknown.
6. Two alternate reads for the opening and closing lines (e.g. more energetic, more intimate).

Do not change the meaning of the script; suggest wording edits separately.

Example (excerpt):
Overall direction: friendly and steady, like explaining something to a colleague over coffee. Around 150 words per minute. Avoid an announcer voice.
Estimated duration: 148 words at 150 wpm = about 59 seconds.
Segment 1 (warm, slower): "Running a small shop / means doing EVERYTHING yourself. //"
Segment 2 (lighter, normal pace): "What if your invoices / could send THEMSELVES?"
Pronunciation: "SaaS" - say "sass"; "Nguyen" - "nwin" (confirm with the person).

When to use it

Use this prompt before recording narration for an explainer video, course module, podcast intro, ad or audiobook sample. It helps when briefing a professional voice actor, when recording yourself, and when generating speech with a text-to-speech tool that responds to punctuation or markup.

The prompt assumes the script is close to final. If the wording is still changing, finish the script first: direction notes tied to specific words become useless as soon as those words change. It also helps to read the script aloud once yourself before running the prompt, so you notice the sentences that are hard to say.

How to customize it

• [WHO WILL READ IT] changes the output. A human needs direction in plain language; a TTS tool needs formatting it can interpret.
• [PROJECT TYPE AND AUDIENCE] set the energy level. An ad and a meditation app need very different reads.
• [DESIRED TONE] works best as a comparison to a real situation (“like a helpful colleague”).
• [TARGET LENGTH] lets the model check timing; 130 to 160 words per minute is a typical range for clear narration.
• [PRONUNCIATION] should include anything a stranger could get wrong, especially people’s names.
• [MARKUP]: if your tool supports SSML or similar tags, say so; otherwise punctuation and line breaks do the work.

Why it works

Most weak voice-overs come from a script that was written to be read silently. Segmenting it and marking stress and pauses forces decisions about rhythm before recording, which saves retakes. Timing based on words per minute catches scripts that are too long for their slot.

Alternate reads for the opening and closing give options for the lines that matter most, and keeping wording edits separate protects approved copy.

Tips and variations

• Record or generate the first segment, listen, and adjust the overall direction before doing the rest.
• Get consent before cloning anyone’s voice; read our guide to text-to-speech and voice cloning.
• Write the script first with the video script prompt.
• Confirm name pronunciations with the people themselves, not only with the model.