A podcast that makes a listener reach for the volume, then startles them with the next episode, has a mastering problem. The words may be clear and the voice may be good, but the listening experience asks the audience to keep fixing the level for you.
Vois makes loudness and mastering part of the same production workflow as the script, selected voice, and export. That matters because a narration is not finished when it has been generated. It is finished when a listener can hear it comfortably in the place where it will actually play.
LUFS, short for Loudness Units relative to Full Scale, is a way to measure perceived loudness over time. It is more useful than looking only at the tallest peak because two files can peak at the same level and still feel very different in volume.
Start with the listening destination
Master for the destination, not for an abstract idea of "loud enough." Streaming platforms and audiobook distributors can normalize playback, and their treatment can change. Check the current requirements of every place you plan to publish, then use a separate approved master when the needs conflict.
These are common destination ranges to confirm before you export:
| Destination | Typical target | Production note |
|---|---|---|
| YouTube and Spotify | Around -14 LUFS | Check the finished video or episode in the destination player. |
| Apple Podcasts | Around -16 LUFS | Compare the final episode with a relevant reference show. |
| ACX audiobook delivery | -23 to -18 dB RMS | Confirm ACX's current peak, noise-floor, file, and narration rules before release. |
Standard ACX uploads do not accept arbitrary externally generated AI narration. ACX's limited Voice Replica beta is a separate, invitation-only workflow. Resolve the exact narration pathway before mastering because a technically correct file can still be ineligible.
Understand what LUFS does and does not tell you
Peak level measures the highest point in an audio signal. LUFS estimates how loud the whole program feels to a listener. That distinction matters for spoken audio, where frequency balance, pauses, vocal energy, and dynamic range all shape the experience.
Two narrations can both avoid clipping while one sounds quiet and the other feels tiring. One may have sharp consonants, too much low-frequency weight, or aggressive compression. LUFS helps you compare overall loudness, but it cannot decide whether the voice is intelligible, natural, or appropriate for the material.
Use the number as a production check, then listen. Headphones can reveal clicks and harshness. A phone speaker can reveal whether consonants disappear. The actual host player can reveal a problem neither of those tests catches.
Use a consistent mastering path in every Vois project
Vois's audio export applies a mastering path that includes loudness normalization, de-essing, EQ, and limiting. Keep the destination choice and the approved export version in the project record so a revised chapter or episode does not quietly leave with a different sound.
The order matters:
- EQ can remove unwanted low-end weight and clarify speech when needed.
- De-essing reduces harsh sibilance without flattening every consonant.
- Compression and limiting control unusually loud moments and protect against clipping.
- Loudness normalization brings the finished program toward the chosen delivery target.
This is not a recipe to apply blindly. A voice that is already sharp may need less high-frequency emphasis. A quiet, intimate narration can become lifeless if the dynamic range is squeezed too hard. Use the master to make the listener's job easier, not to force every voice into the same texture.
Ask an agent for an export plan, then approve the master
An agent can prepare the repetitive parts of an export workflow, such as grouping approved scripts by destination and listing the checks each release needs. It should not decide what a listener receives.
Prompt your agent: "From these approved Vois projects and distribution requirements, prepare an export plan. Group each item by destination, preserve the approved script and voice revisions, flag conflicting loudness or format requirements, and list the playback checks needed for each master. Do not export, alter mastering choices, or publish files until the producer approves the plan."
Use the review flow:
- The producer confirms the current distribution requirements and the project version to master.
- The editor approves the selected export path and generates one representative master.
- Reviewers listen on headphones, a small speaker, and the intended playback environment, then compare it with a relevant reference.
- The release owner approves the named master, captions or transcript where relevant, and the destination upload.
This keeps a useful agent task on the planning side of the line. The human team still listens for distortion, pumping, buried words, or a pacing problem that a loudness measurement cannot identify.
Common mastering mistakes are often review mistakes
Over-compression can make a narration feel flat and tiring. A master that is too loud can trigger platform normalization or leave too little room for the program's own dynamics. A master that is too quiet forces the listener to compensate. These are listening problems before they are meter problems.
Another common mistake is exporting once and assuming the job is complete. When the script changes, recheck the replacement in the full program. A new name may introduce sibilance, a longer line can change the cadence, and a revised scene can alter the balance against music or effects.
Vois helps because the script, takes, and export decisions live together, so a specific approved change can be traced to the master that followed it. That is more dependable than trying to recreate an old process from an unnamed file called final.
Keep the mix and master decisions separate in the review notes. The mix decides how narration relates to music, effects, and other speakers. The master prepares that finished program for a delivery target. If words are buried under music, changing the loudness target is not the right fix. Return to the mix, protect the spoken information, then master the approved result again.
This distinction is especially helpful for revisions. A new narration take may need a small mix adjustment even when the destination and mastering path remain unchanged. Review the complete program before approving the replacement master.
Let the listener have a stable experience
Good mastering is mostly invisible. The listener notices the story, instruction, or conversation instead of the volume. That is the standard worth aiming for.
Get started with Vois audio export, then download Vois to master one approved project in its actual listening context. See pricing for current plan details.
The Vois Team