Prepare AI voiceover by writing spoken text, checking pronunciation and testing pacing in the chosen voice tool. A script-writing request produces text; a text-to-speech workflow turns that text into audio.

Example brief and prompt

We need a short library-workshop introduction. Confirmed details: Saturday at 10:00, Room 4, bring a laptop. Booking is required; the link is pending. The script should sound calm and use words that are easy to pronounce.

Write a short spoken introduction using only these facts:
Saturday at 10:00, Room 4, bring a laptop.
Booking is required; link pending.
Use plain sentences and spell numbers as they should be spoken.
Return the narration separately from production notes.
Do not add music instructions inside the spoken script.

Example narration

Join the library workshop on Saturday at ten in the morning, in Room Four. Bring a laptop. Booking is required. The booking link is pending confirmation.

Production notes: pause between the location and equipment sentence; confirm the exact date before publication; keep delivery calm. These notes are directions for production, separate from the words to be spoken.

Hand off to a voice tool

Select a voice you have permission to use. Paste the narration into the text-to-speech input and apply the supported controls for the selected model. ElevenLabs’ speech best practices describe model-dependent pacing, pronunciation and pause handling. Avoid assuming one set of tags works across every voice model.

ElevenLabs’ Instant Voice Cloning requires rights and consent for the voice. Its Professional Voice Cloning is limited to your own verified voice; someone else’s consent does not make that workflow available. Our ElevenLabs Quick Start covers the product steps.

Listen and revise

CheckAction
Words match approved scriptListen while reading the text
Numbers and names are clearExpand ambiguous notation or use supported pronunciation controls
Pacing fits the videoMeasure actual audio duration and adjust text or supported delivery settings
Production notes stay separateRemove any accidentally spoken stage directions

If “10:00” sounds unnatural, replacing it with “ten in the morning” is a script edit you can compare directly. If the clip is too long, shorten a sentence and regenerate the affected passage instead of assuming a universal words-per-minute rate.

For practice, prepare a second script with a name or acronym and document the pronunciation you need. Keep the spoken facts aligned with the script and storyboard.

Listen to a speech rehearsal

Example rendered with Flite speech synthesis: 9.61 seconds. Listen for the spoken time and room number, then compare the pacing with your intended video.

Read the audio transcript

Join the library workshop on Saturday at ten in the morning in Room Four. Bring a laptop. Booking is required. The booking link is pending confirmation.

Download the rehearsal audio. Measure the output from your selected voice workflow separately, then adjust its script and supported settings.