Lightly, and for sound, not for the numbers. Corrective EQ that tames boom or harshness, a de-esser that softens harsh s sounds and gentle compression that keeps the natural gap between your words and your pauses are all fine, and often good. What must wait for mastering are the mastering moves: a limiter, loudness normalisation to the ACX target, heavy compression that flattens the performance, one-click "enhance" processing, and noise reduction pushed to reach ACX's -60 dB noise floor, which mastering reaches by lowering the noise in the pauses instead. The test is simple. If you are doing it to hit an ACX number, stop: that is mastering's job. If it simply sounds better and the pauses stay quiet next to the words, keep it.
Mastering is the final pass that sets the loudness, holds the peaks under the limit and lowers the noise in the pauses. ACX's production blog, written in 2014, points the same way: a good raw recording leaves the headroom to master "without needing to compress the signal heavily", and its limiter step sits in the mastering chain.[1] If you master the file yourself, the mastering moves belong in that one pass, applied once.
Every app setting and quote on this page was checked against its official source on 30 October 2026.
What counts as processing before mastering?
Two kinds, and the difference is what each one is for.
Processing for sound: fine, used lightly
- EQ (equaliser): boosts or cuts ranges of frequencies. Used correctively, it tames a boomy low end or a harsh edge.
- De-esser: turns down sibilance (the hiss of s sounds) only at the moments it occurs.
- Gentle compression: a compressor turns loud passages down so the whole file can be raised.[2] Set gently, it evens out the voice and leaves the pauses well below the words.
Mastering moves: leave them for mastering
- Limiter: a compressor that lets no peak pass a set ceiling.[2]
- Loudness normalisation: raises or lowers the whole file to hit a target level, such as ACX's loudness window.
- Heavy compression: the same compressor pushed until quiet and loud passages sit almost on top of each other.
- Heavy noise reduction: subtracts a profile of the background noise from the whole recording, voice included, pushed hard enough to reach the noise floor limit.
- "Enhance" features: one-click voice processing. Descript's Studio Sound, for example, "enhances spoken voice by reducing background noise, echo, and other distractions".[3]
Editing is different again. Cutting a fluff, a cough or a long gap changes what is in the recording, not how your voice sounds, and it belongs before mastering.
Why do mastering moves before mastering make the file worse?
- It cannot be undone. Once heavy compression or a limiter has narrowed the gap between your words and your pauses, mastering cannot put it back. It works with the performance it is given.
- It lifts everything that is not your voice. The quiet moments are where your breaths, mouth sounds and room noise live, and heavy compression raises them. ACX's blog warns that compression can reduce the dynamic range of a vocal recording "if used improperly".[1] Those pauses are also where the noise floor is read, and ACX requires it to be no higher than -60 dB RMS (RMS is the average level).[4] Pauses pushed up towards the words leave mastering less room to get there.
- It gets done twice. A limiter on the recording and another in mastering squeeze the same peaks twice. Each pass adds its own side effects, and the second cannot remove the first.
Noise reduction subtracts from your voice as well as the pauses. At heavy settings, Audacity's guide warns, your voice "may get wine-glassy or honky".[5] The noise floor guide explains why.
What does an over-processed chapter look like?
One chapter a narrator sent us to look at shows the pattern. Before mastering, it already carried the marks that heavy compression and a limiter leave behind.
- Almost no movement. There was only 2.1 dB of movement between its quieter and louder passages.
- Peaks pinned to one ceiling. Its loudest peaks all sat within a decibel of one ceiling, the trace a limiter leaves.
- Pauses close to the words. The pauses sat only 11 dB below the speech, filled with breaths and a music bed under the narration.
- A lossy export already in the chain. A sharp cut in the highest frequencies showed an earlier lossy export (a format such as MP3 that discards some of the sound).
Mastering brought its loudness and peaks inside ACX's limits and lowered the noise floor by 19.8 dB, but the chapter still failed the noise floor. The main cause was the music bed, which sat inside the short pauses and under the words, already squeezed up towards the voice. The music under narration guide covers that side. The fix sits upstream: leave the limiter and the heavy compression to mastering, export the voice on its own, and master once.
What should you send for mastering?
A clean, edited WAV of the voice alone, with any processing you chose for sound and nothing aimed at the spec. In order:
- Record clean, at a sensible level. A raw recording is meant to be quieter than the finished file. ACX's blog describes a well-recorded file before mastering as having an RMS no greater than -28 dB and peaks no higher than -12 dB.[1] Loudness is mastering's job: the audiobook loudness guide explains the window the finished file lands in.
- Keep an untouched copy. Export a WAV of the raw reading and keep it somewhere safe before you edit anything.
- Edit. Cut fluffs, retakes, coughs and long gaps. Clear the extra sounds ACX lists: plosives, microphone pops, mouse clicks and excessive mouth noise.[4]
- Process for sound, lightly, if it helps. Corrective EQ, a de-esser or gentle compression are fine. Listen to a pause afterwards: it should still sit well below the words.
- Leave the spec to mastering. No limiter, no loudness normalisation, no heavy noise reduction, no "enhance". The list below shows where apps and computers hide them.
- Export the voice on its own, as WAV. Mute any music or sound effect tracks, so mastering works on the voice alone. WAV is lossless and MP3 is not, and your final ACX file will be an MP3 anyway.[4]
- Master once, then measure. Measure the mastered file before you upload. The free ACX check reads RMS, peaks and noise floor in your browser.
Where do apps add processing you may not notice?
GarageBand switches one setting on by default, and other apps keep effects in places that are easy to miss. Keep what you chose for sound, and switch off anything aimed at the spec. Each entry comes from the maker's own documentation.
GarageBand (Mac)
- What it adds: The Export projects at full volume setting raises the export to the highest level without distortion.[6] Apple's guide for GarageBand 10.4 lists it as the default.[7] Export also trims any silence at the start and end of the project.[8]
- Before you export: Untick it in GarageBand > Settings > Advanced.[6] Export with Share > Export Song to Disk and choose WAVE.[8] Apple's guide lists no setting for the trim, so check the ends after mastering: ACX recommends 1 to 5 seconds of room tone (the quiet sound of your room) at each.[4]
Descript
- What it adds: Studio Sound, when switched on, reduces noise and echo in the voice.[3] Export has a Normalize volume setting.[9]
- Before you export: Toggle Studio Sound off in Audio Effects.[3] Set Normalize volume to Off under Advanced.[9] To export the voice alone, open the Sequence Editor, select the voice track and click its S (solo) icon, exit the editor, then export: Descript exports only the soloed track, as a mono file.[10]
Windows
- What it adds: Audio enhancements on the microphone input.[11]
- Before you export: Start > Settings > System > Sound, choose your microphone under Input, then Device Properties > Additional device properties. Tick Disable all enhancements on the Enhancements tab, or untick Enable audio enhancements on the Advanced tab, depending on your manufacturer.[11]
Adobe Audition
- What it adds: A multitrack mixdown carries the volume, pan and effects settings routed to the mix track.[12]
- Before you export: Switch off any limiter or loudness effect on the voice track and the mix track, keeping only what you chose for sound, then use File > Export > Multitrack Mixdown.[12]
REAPER
- What it adds: A master mix render includes the master track's FX, and a stems render is taken after each track's FX, fader and pan.[13]
- Before you export: Bypass any limiter or loudness FX before rendering, or use File > Consolidate/Export tracks to leave all track FX out, including the ones you chose for sound (effects on a single clip still go in).[13]
Audacity
- What it adds: Realtime and master effects leave the waveform untouched while you work, but the stack is applied when you export.[14]
- Before you export: Turn any limiter or loudness effect off with its power button, or remove it from the stack.[14]
What if you master the file yourself?
Then the rule is "process once": the mastering moves happen in one pass. The Audacity mastering guide chains noise reduction, a compressor, loudness normalisation and a limiter. What matters is where those effects sit.
- In one pass, on the edited recording. Not on the track while you record, and not again on a file that was already mastered. The Audacity guide's macro troubleshooting goes further for its own macro: run it on unprocessed, edited files.
- Once per file. A file you master yourself and then send to a mastering tool or an engineer gets mastered twice. Send them the edited WAV, without the mastering moves.
A mastering tool expects the same starting point: what chapterpass does lists each step chapterpass takes and the editing it leaves to you.
Common questions
Can mastering undo compression that is already in the file?
No. Mastering can set the loudness, hold the peaks under the limit and lower the noise in the pauses, but it cannot restore the gap between words and pauses once heavy compression has narrowed it. If you still have the recording from before that compression, master that instead.
My raw recording sounds quiet. Should I boost it before mastering?
Should I use noise reduction before sending my file for mastering?
Lightly, for a specific problem such as a hum, yes. Pushed hard to reach -60 dB, no: it subtracts from the voice too, and mastering lowers the noise in the pauses for you. The noise floor guide shows what heavy settings do to the voice.
Ready to submit?
Run the free ACX check to see exactly where your file stands, or let chapterpass master it to spec in one step.