The narration is approved. Then the editor asks which file is current, why the music is attached, and whether the opening silence belongs there. Delivering voiceover files is part of the production work, not an administrative chore after it.
A useful handover lets someone import the right audio without interpreting your folder. Agree the specification before exporting, give every asset a stable identity, and explain exactly what changed. Your client should spend their attention on the edit, not on solving a filename puzzle.
What should you ask before exporting voiceover files?
Ask where the audio goes next. A file headed into a video timeline has different needs from a finished listening copy. Find out whether the recipient wants individual scenes, a continuous read, separate dialogue tracks, or a finished mix.
Confirm format, sample rate, bit depth where applicable, channel layout, loudness requirements, and the expected start point. Ask whether clips should include leading space or begin close to speech. Record the answers in the delivery brief instead of relying on a remembered chat message.
If the client cannot provide a specification, propose one and get agreement before the final export. Don't guess that the destination name tells you everything. A podcast editor may want an unmastered narration source even though the eventual episode needs a finished podcast export.
The voiceover production workflow ends at the recipient's import, not at your export button. A short sample delivery can reveal a mismatch while it is still cheap to change the plan.
How should you name voiceover files for an editor?
Use names that identify the project, asset, language, and revision without requiring the folder to explain them. An example convention is atlas_intro_en-GB_r03.wav. That's an example, not a required format; match the client's existing convention when there is one.
Keep the asset ID stable across revisions. If the words change but the scene's purpose stays the same, don't rename it to the new opening phrase. The editor needs to recognize the replacement. Keep approved script IDs in the handover sheet beside the filenames.
Use a consistent ordering scheme if files must sort into sequence. Decide whether a number means scene order or asset identity, and don't switch halfway through production. Reordering a video shouldn't accidentally make an old clip look like a new one.
Avoid final, final_new, and final_actual. A revision identifier and approval note tell a clearer story. Put alternatives in a clearly marked folder instead of mixing them with the approved set, even if the client once asked to hear both.
Which audio format should you deliver for each destination?
For editable source audio, propose WAV unless the recipient requests something else. It is a straightforward choice for handing narration into an editing workflow. Confirm the actual encoding and technical settings rather than assuming a filename extension tells the whole story.
Use MP3 or AAC for a convenient review copy when the client accepts those formats. They are useful when listening and sharing matter more than further processing. Don't treat a review copy as the master simply because someone approved its performance.
FLAC can be useful for lossless storage or transfer when the receiving workflow supports it. Confirm import support before choosing it to save space. An inconvenient file that preserves everything still creates unnecessary work for an editor who cannot open it.
| Delivery purpose | Practical starting choice | Confirm before sending |
|---|---|---|
| Editable narration | WAV | Encoding, sample rate, channel layout |
| Listening review | MP3 or AAC | Playback support and clear review labeling |
| Lossless archive | WAV or FLAC | Archive policy and future tool support |
| Finished release | Destination specification | Current requirements and final mix approval |
Vois audio export includes MP3, WAV, FLAC, and AAC, with destination presets. Choose the preset for the actual deliverable, then inspect the file. A preset name doesn't establish a platform's permission to publish AI narration; the client must check the platform's current rules separately.
Should you deliver separate stems or a single mix?
Deliver separate components when the editor needs control over their balance. A narration-only track lets them adjust speech without changing the music. A finished mix is appropriate when the combined presentation is itself the approved deliverable.
Define “stem” in the brief. One person may mean all narration together; another may want each speaker separately. Name the requested components explicitly, including whether effects and music are part of the handover and whether you have permission to share those source assets.
For full-length stems, use a shared start point and aligned duration so they line up when imported together. For individual scene clips, provide the scene mapping instead. Don't strip silence from some aligned files while leaving it in others.
Keep a narration-only version when supplying a reference mix with music. Explain which file is for editing and which demonstrates the intended balance. Only promise exports you have actually prepared and checked; don't imply that every component will be separated automatically.
What loudness target should a client voiceover use?
Use the client's agreed specification for the deliverable, and distinguish a narration source from a complete programme. Integrated loudness describes the level across the measured audio. A final mix target isn't automatically the right target for every short scene clip or separate stem.
For a specific broadcast example, the European Broadcasting Union's R 128 recommendation, revised November 2023, gives a programme target of -23.0 LUFS and a maximum true peak of -1 dBTP during linear audio production. It also says to measure the programme in its entirety, not only the speech. Those figures are not a universal prescription for online voiceovers.
Ask the editor whether they want clean source narration with room for the mix or a mastered, ready-to-publish programme. If they will add music and effects, they need to evaluate the combined result. Normalizing every supplied stem independently to a programme target can undermine the intended balance.
Use Vois mastering controls deliberately and listen after processing. The audio mastering guide explains the concepts; the delivery brief remains the authority for this job. Report measured loudness and peak values only if you've actually measured the delivered file.
How do you number revisions without confusing the client?
Increment the revision when the delivered audio changes, and identify exactly which assets replace earlier ones. Keep the same base ID so the relationship is obvious. Don't overwrite a sent file under the same name and expect everyone to refresh their download.
Attach a compact change list: asset ID, prior revision, new revision, and the reason. “Corrected brand pronunciation” is more useful than “updated.” Mention duration changes because they may affect picture sync even when the words are almost identical.
Send replacement-only packages only when the client expects them. Otherwise deliver a complete current set with the changes clearly marked. In either case, state whether unchanged files remain valid.
Keep an internal archive of previous approvals. The client's current folder should make the right choice obvious, but your own record should preserve what was sent and why. Never label an unreviewed replacement as approved just because an earlier take passed.
What belongs in the final voiceover handover note?
Include the project and script version, delivery date, file inventory, approved revision, technical format, processing status, and any unresolved exceptions. State the start-point convention and whether the package contains source tracks, a mix, or both.
Add pronunciation decisions and usage information relevant to the job. If a correction remains pending, identify the affected file instead of hiding the caveat in a general sentence. Our export presets guide can help explain the export choice, but the note should describe what this package actually contains.
Before sending, open the exported files from the delivery folder, not the project browser. Check their starts, ends, names, and representative transitions. Confirm that the inventory matches the folder and that the sharing permissions let the intended recipient download it.
Deliver audio the editor can trust before they press play.
The Vois Team