The missing word always seems obvious after you publish. A listening test before publishing gives that word somewhere else to be caught: in a repeatable review of the file your audience will actually receive.
Don't approve the timeline from memory. Export the intended deliverable, close the editing view, and open that file in an ordinary player. We recommend the protocol below as an editorial routine, not a scientific claim about attention spans or an assurance that every device will sound identical.
What should I prepare before starting a listening test?
Have the exported file, the approved script, a place to record timestamps, and the destination's current delivery requirements. Confirm that the filename belongs to the revision you intend to publish. An excellent review of yesterday's export is still the wrong review.
Keep a reference piece available if you are maintaining an existing series. Compare similar speech at a stable playback volume rather than turning your new episode up until it feels impressive. Write down problems as location, symptom, and intended fix. “Sounds weird” is hard to hand off; “the name after the introduction loses its final syllable” is actionable.
Use audio export to create the intended format and preset, then review the result. A preset is a starting point, not proof of acceptance by a platform. Content policies, including rules for synthetic narration, remain separate from audio-file requirements.
Why should I check the first 30 seconds separately?
Treat the first 30 seconds as an opening checkpoint, not a researched deadline for listener attention. We suggest this short window because it is easy to repeat after revisions and long enough to inspect how the narration begins.
Start from silence. Does the first word arrive intact? Is there a startling level jump? Can you tell who is speaking and what the piece is about without seeing the title? Listen for an introduction that makes sense only because you already know the script.
Check the transition into the main content as well. Music may sit comfortably beneath a short title but cover consonants when the explanation begins. A pause that feels dramatic in the editor may sound like playback has stopped. The opening should invite attention without making the listener troubleshoot the file.
If you change the opening, repeat this checkpoint on the new export. Do not approve it solely because the replacement sentence sounded good in isolation.
What should I listen for on speakers, earbuds, and a phone?
Use each device to ask a different question. Speakers help you hear how speech and music share space in a room. Earbuds can make small joins, distracting mouth-like sounds, and sharp consonants easier to notice. A phone speaker gives you a practical check that meaning does not depend on a full bass response.
Keep the volume comfortable. This is a listening test, not a contest to expose the quietest artifact at an unrealistic level. If a device has unusual enhancement settings, note them. Don't mistake a playback effect for a fault in the export.
| Playback device | Main question | A useful failure signal |
|---|---|---|
| Speakers | Does the voice remain clear in the mix? | Music competes with ordinary sentences |
| Earbuds | Are edits and consonants comfortable? | Clicks, sharp sibilance, abrupt joins |
| Phone speaker | Can I follow the message without strain? | Names or instructions disappear |
You don't need identical sound across devices. You need intelligible speech and no distracting surprises. If a passage fails everywhere, repair the source or mix. If it fails only on one device, compare another file there before deciding what the export needs.
Should I listen at 1x or 1.5x speed?
Listen to the complete export at 1x first. That is the intended performance, and it exposes pauses, phrasing, and emotional shifts that faster playback can hide. Use the approved script for an accuracy pass, then look away for difficult sections to see whether the meaning survives without visual help.
Use 1.5x as an additional stress test for dense explanations, names, abbreviations, and transitions. These speeds are our suggested review settings, not a claim that everyone listens this way. If words blur together, consider shorter sentences or clearer separation between ideas rather than simply slowing the entire project.
Don't force every style to suit accelerated playback equally. A meditation, a dramatic scene, and a technical briefing have different intentions. The faster pass can reveal a weak join without becoming the artistic standard for the piece. Our speed and pacing guide explains how to keep rhythm tied to meaning.
How do I check loudness against the destination?
Read the destination's current specifications, then measure the exported file with a suitable loudness meter. Perceived loudness and peak level are different questions: a file can have an acceptable peak while its overall speech level is unsuitable for delivery.
Do not copy a broadcast target into every online project. The European Broadcasting Union's Loudness in Streaming recommendation, published November 21, 2023, addresses streaming as a distinct context. Its accompanying recommendation notes the broadcast production target of minus 23 LUFS while discussing streaming delivery separately. That number is context, not a universal upload instruction.
Vois provides destination-oriented export presets, but verify the actual result against the requirements that apply to your file. If you change music, EQ, or level after measurement, measure again. Consult the export presets guide for the production workflow, then use the destination itself as the authority for acceptance.
Avoid fixing a mix problem by raising everything. If music masks speech, change the balance. If the voice is harsh, investigate the source and appropriate EQ or de-essing. Loudness adjustment alone cannot restore a missing syllable or make an unclear sentence informative.
What pass or fail checklist can I reuse?
Give each line a pass, fix, or not-applicable decision. “Probably fine” belongs in the fix column until someone has listened. For podcast production, keep this checklist alongside the episode script so another editor can see exactly what was approved.
- Identity: the filename, revision, title, and intended destination match.
- Opening: the first word is intact and the introduction is understandable.
- Accuracy: names, numbers, instructions, and the ending match the approved script.
- Delivery: phrasing and pauses support the meaning at normal speed.
- Edits: no missing words, distracting joins, clicks, or accidental silences remain.
- Mix: speech remains understandable on the checked playback devices.
- Delivery specification: measured loudness, peaks, format, and other required properties are acceptable.
- Final file: the reviewed export is the exact file selected for publication.
Treat factual errors and missing speech as failures, even if the rest sounds beautiful. Cosmetic preferences can be documented separately, but do not let a long wish list hide a blocking defect.
What should I do after finding a problem?
Fix the smallest relevant cause in the project, not an unrelated setting that happens to disguise it. Correct a word in the script, repair a join in the timeline, or adjust a competing music passage in the mix. Regenerate only the affected material when possible, preserving approved work around it.
Export again and check the repaired passage with its surrounding context. Then repeat any checks the change could affect, including loudness after mix changes. Keep the rejected file clearly separate from the approved one so the publishing step cannot quietly undo your review.
Publish the file you checked, not the one you meant to check.
The Vois Team