# Audio-First Video Production: Build a Voiceover Timing Map

> Build a voiceover timing map for AI documentaries: approve narration, measure beats, assign visual purpose, direct shots, and control revisions.

Updated: 2026-07-13
Audience: AI documentary producers, editors, and YouTube teams planning picture against an approved voiceover
Canonical: https://onira.studio/guides/audio-first-video-production

## Key takeaways

- Use the accepted audio file, never word count, as the clock.
- Give every beat exact timecodes, one viewer outcome, and a visual budget.
- A changed voice track must explicitly invalidate dependent picture and finishing work.

## 1. Accept and version the narration master

Review factual wording, pronunciation, pace, emphasis, silence, clipping, and completeness in the exact file intended for the cut. Record the script revision, voice, language, file identity, reviewer, and approval date. A new performance is a new timing contract even when the filename remains unchanged.

The master should be decoded and measured from the media itself. Word count and text-to-speech estimates are planning hints; they cannot own shot duration, captions, or final assembly.

- Exact accepted script and audio identity.
- Pronunciation and performance approval.
- Measured duration from decoded media.

## 2. Segment the track into editorial beats

Split at changes of claim, example, place, time, scale, subject, emotional pressure, or act. Keep intentional pauses as real ranges. Preserve word alignment for captions when available, but choose beat boundaries editorially rather than at every sentence or punctuation mark.

Give each beat a stable ID, audio in and out time, approved spoken text, source or uncertainty note, and one sentence describing what the viewer must understand, notice, or feel by the end.

- Stable beat ID and precise time range.
- One viewer outcome per beat.
- Claim and uncertainty boundary attached to picture planning.

## 3. Allocate a visual budget and continuity boundary

Divide each measured beat into one or more visual jobs: orient, make concrete, compare, reveal consequence, or transition. Cut when the editorial idea changes, not because a fixed interval elapsed. A dense line may need restrained picture; a long explanation may need orientation, human detail, and consequence.

For every shot, separate appearance from motion. Record start and end state, geography, screen direction, people, wardrobe, objects, weather, and the transition into the next accepted shot. This turns the map into a usable contract for generation and continuity review.

- Exact shot budget inside the beat duration.
- Separate image and motion direction.
- Explicit transition and continuity state.

## 4. Finish and revise from the same clock

Captions should align to the accepted take. Score should follow real act durations. Sound effects should fit the scene without masking words. The canonical timeline should reference the same narration version and beat IDs used by picture planning.

When the script or performance changes, calculate and invalidate the affected scope: source record, beat timing, shot durations, trims, captions, score, effects, and neighboring transitions. Preserve accepted unaffected work, then rerun one uninterrupted sequence review after repair.

- One audio version across picture and finish.
- Bounded invalidation after a change.
- Normal-speed review before technical diagnosis.

## Publication checklist

- The exact narration file is accepted, measured, and versioned.
- Each beat has an ID, time range, spoken text, and viewer outcome.
- Each shot has a purpose and fits its measured budget.
- Appearance, motion, transitions, and continuity are separate fields.
- Captions, score, effects, and render use the same clock.
- Revision rules preserve unaffected accepted work.

## Sources

- [Runway guidance for creating longer films](https://help.runwayml.com/hc/en-us/articles/26871350018835-How-to-create-longer-videos-and-films)
- [Google Veo video-generation documentation](https://ai.google.dev/gemini-api/docs/video?hl=en)

## Questions

### What does audio-first video production mean?

For narration-led work, it means the exact approved voice track is measured and converted into a beat-level timing map before final visual architecture is generated.

### What belongs in a voiceover timing map?

Use beat IDs, audio in and out points, approved text, viewer outcome, source or uncertainty note, visual jobs, shot ranges, transitions, continuity state, and acceptance status.

### What happens when the narration changes?

Create a new audio version and explicitly invalidate affected timing, picture, captions, score, effects, and transitions. Preserve accepted work outside the changed scope.

## Product boundary

- Onira delivers a final MP4; it does not upload or schedule posts on YouTube or social platforms.
- Onira provides a reviewable production workflow; creators remain responsible for approving the story, facts, rights, disclosure, and final publication.
- Director chat is limited to regenerating one selected PREVIEW timeline video clip; other available Studio controls are separate direct actions.
- Creators must review facts, sources, rights, realistic-synthetic-media disclosure, and platform policy before publishing.
- Onira does not guarantee YouTube monetization, reach, factual accuracy, or legal clearance.