How do you generate narration that matches the approved script?

How do you generate narration that matches the approved script?

Narration should be generated from a specific approved text, not an editor buffer that may change while the request runs.

Freeze the input for the attempt

Save the script version and selected voice settings with the generation job. Keep stage instructions separate from spoken text. If a provider requires chunking, split at sensible boundaries and preserve ordering. Do not assume separately generated chunks will have identical pacing or tone.

Review the resulting speech

Listen for names, numbers, abbreviations, missing words, and abrupt transitions. Check that the output includes the full script without an unexpected introduction or ending. Measure actual duration before building the video timeline. If the creator changes text, create a new narration version and clearly identify dependent captions and renders as outdated. Test a script with a difficult name and a correction after approval.