How to master an audiobook in Audacity: peak, RMS and noise floor
To master an audiobook for ACX in Audacity, cut the low rumble, set the RMS loudness to −20 dB and hold the peaks with a limiter at −3.5 dB. Then export a 192 kbps constant-bit-rate MP3 and measure that file. It passes ACX’s technical check when peaks stay below −3 dB, RMS sits between −23 and −18 dB and the noise floor stays below −60 dB.
This guide covers the three ACX numbers, one chapter in Audacity 4 and a whole book with the Audacity 3 macro. It also covers fixes for the noise floor, why a file that passes ACX Check can still fail review, and voice plugins that help.
Tested on Audacity 4.0.1 on macOS Sequoia and Windows 11, and on Audacity 3.7.9, the version MuseHub installs as Audacity 3, on Windows 11; the screenshots are from Audacity 4.0.1. Shortcuts are written for Windows; macOS uses Cmd where Windows uses Ctrl.
Audacity and voice tools on MuseHub
The three numbers ACX checks
ACX (Audiobook Creation Exchange), the platform where narrators publish audiobooks to Audible, runs every uploaded file through an automatic check before a person listens to it. Three of its numbers are about level.
| Requirement | ACX limit | What it measures | What moves it in Audacity |
|---|---|---|---|
| Peak | Below −3 dB | The single loudest sample in the file | The Limiter |
| RMS | Between −23 and −18 dB | The average level of the whole file, which works out to how loud it sounds | Loudness Normalization in RMS mode |
| Noise floor | Below −60 dB RMS | How loud the file is when you stop talking, from the room, a computer fan or microphone hiss | The room and the microphone first, then Noise reduction |
ACX’s submission requirements also set the format. Each file is an MP3 at 192 kbps or higher with a constant bit rate (CBR) and a 44.1 kHz sample rate. It holds one chapter or section, runs no longer than 120 minutes, and opens and closes with 1 to 5 seconds of room tone. All files in a book are mono, or all are stereo, and they sound consistent from chapter to chapter.
These numbers are the first gate. Editing errors, extraneous noise and dynamics problems are left to ACX’s quality team, which listens after submission.
What you need
- Audacity 4 for one chapter at a time. It is free, through MuseHub on Windows and macOS or from audacityteam.org, which also has a Linux build.
- Audacity 3 for a whole book. Audacity 4.0 has no macros. MuseHub lists the older version as Audacity 3; both are free and install side by side.
- The ACX Check plugin. It reads peak, RMS and noise floor against ACX’s limits. Get it from the Audacity plugins site and install it as Step 5 shows. Neither Audacity 4 nor Audacity 3 ships with it.
- A raw recording of one chapter, with 1 to 5 seconds of room tone at the start and the end.
Back up the raw read first
Before any effect, export the unprocessed chapter as a 16-bit WAV file and store it away from the computer. The Audacity Support guide to audiobook mastering puts this first, so a damaged edit never costs a second read.
In Audacity 4, choose File ▸ Export audio… (Ctrl+Shift+E), then Export to computer. Set Format to WAV (Microsoft) and Encoding to Signed 16-bit PCM.
Master one chapter in Audacity 4
The first two steps use Effect-menu effects, which rewrite the waveform. The limiter runs live in the Master section of the realtime effects panel, which processes playback and the export but not the waveform, so the check reads the exported file.
Step 1: Cut the low rumble
Traffic, footsteps and air conditioning put sound below the voice, and ACX Check counts that sound as noise in every pause.
- Open the EQ. Press Ctrl+A to select the chapter, then choose Effect ▸ EQ and filters ▸ Filter Curve EQ.
- Load the speech preset. Open the preset list at the top, which reads Default preset, and choose Low rolloff for speech. The Support guide uses the same preset in Audacity 3, where it leaves 100 Hz and above untouched and cuts everything below 50 Hz completely.
- Preview and apply. The voice should sound the same, with less rumble in the pauses.
High-pass filter, in the same EQ and filters group, does the job too. Its default cutoff of 1000 Hz would thin the voice badly. Set Frequency to about 80 Hz, inside the 60 to 100 Hz the Audacity 4 manual suggests for speech, and Roll-off to the steepest option, 48 dB per octave.
Step 2: Set the RMS loudness to −20 dB
Loudness Normalization moves the whole selection so its average lands on the target. −20 dB sits in the middle of ACX’s window, with room on both sides.
- Open the effect. Keep the chapter selected and choose Effect ▸ Volume and compression ▸ Loudness Normalization.
- Switch the measure to RMS. The dialog opens on Perceived loudness at −23 LUFS. Set the Normalize dropdown to RMS, and the target changes to −20 dB, the value you want.
- Leave Normalize stereo channels independently off. Treat mono as dual mono applies only to Perceived loudness, so it does nothing here.
- Apply.

Perceived loudness, in LUFS, is the measure podcast and streaming platforms use; ACX checks RMS.
Step 3: Hold the peaks with a limiter on the Master
Raising the average also raises the peaks, often past −3 dB. A limiter stops every peak at its ceiling and leaves the rest of the voice alone.
- Open the realtime panel. Click the track, then press E or click its Effects button.
- Add the Limiter to the Master. Scroll to the Master section at the bottom, click its Add effect button and choose Limiter from the Audacity group.
- Set Threshold and Output both to −3.5 dB. With the two equal, the limiter adds no level and only catches peaks, half a decibel inside ACX’s limit, like the Support guide’s Audacity 3 soft limit.

The waveform still shows peaks above −3.5 dB, because Master effects act on the sound on its way out. How to make audio louder in Audacity explains the Master section.
Step 4: Export a 192 kbps MP3
Choose File ▸ Export audio… (Ctrl+Shift+E), then Export to computer, and set:
- Format: MP3 Files.
- Channels: Mono for a mono voice. Whatever you pick, use it for every chapter.
- Sample rate: 44100 Hz.
- Bit Rate Mode: Constant, then Quality 192 kbps, which Constant selects by default.
- File name: the chapter number and title, so files upload in order.
Step 5: Measure the exported file
Measure the MP3 itself. The limiter exists only in the export, and MP3 encoding moves peaks a little. Answers on the Audacity forum put that change at usually less than 1 dB.
- In Audacity 4. Install ACX Check once. Choose Tools ▸ Nyquist plugin installer and select the downloaded file, then open Analyze ▸ Plugin manager and click Rescan plugins. To measure, import the MP3 into a new project, press Ctrl+A and choose Analyze ▸ Nyquist ▸ Steve Daulton ▸ ACX Check. It shows Peak level, RMS level and Noise floor, each marked Pass or Fail next to the ACX range, such as “RMS must be in range -23 to -18 dB”. If a Nyquist plugin answers “Effect not found” on a Mac, the fix is in the FAQ below.
- In Audacity 4, without the plugin. Analyze ▸ Measure RMS gives the RMS on the whole file and the noise floor on a stretch of room tone. Amplify opens at the gain that would bring the peak to 0 dB, so the peak is that figure with a minus sign.
- In Audacity 3. Install ACX Check once with Tools ▸ Nyquist Plugin Installer…, then open the MP3 with File ▸ Open, press Ctrl+A and choose Analyze ▸ ACX Check. The panel shows peak, RMS and noise floor against ACX’s limits.
- In your ACX account. ACX’s Audio Lab checks uploaded files for RMS, peaks, bit rate and sample rate, but its list leaves out the noise floor.
If a number fails, fix it in the Audacity 4 project and export again.
A whole book: the Audiobook Mastering macro in Audacity 3
A macro runs a saved list of effects in one go. Macros are not in Audacity 4.0 and may come back in a later version, according to the Audacity team. The Support guide offers a ready-made Audiobook Mastering macro for Audacity 3 with the same three settings. There the limiter writes into the waveform, so ACX Check reads the project correctly too.
- Import the macro. Download the macro file from the Support guide, choose Tools ▸ Macro Manager, click Import… in the Manage Macros window and select the file.
- Master a chapter. Open the chapter, press Ctrl+A and choose Tools ▸ Apply Macro ▸ Audiobook-Mastering-Macro. Run Analyze ▸ ACX Check, then export the MP3 with Bit Rate Mode Constant and Quality 192 kbps.
For a folder of chapters, the macro can run unattended:
- Set the MP3 options once. The macro’s Export as MP3 command has no settings of its own and reuses the MP3 options of your last export. So export one file all the way with File ▸ Export Audio…, Format MP3, Bit Rate Mode Constant and Quality 192 kbps. Cancelling the dialog does not keep the options.
- Edit the macro. In the Macro Manager, insert Select All above the first step, since a macro run on files starts with nothing selected, and Export as MP3 below the last.
- Apply it to files. Choose Apply Macro to Files… and select every chapter WAV in the folder. The MP3s land in a folder named macro-output.
- Check each MP3 with ACX Check before you upload.
When the noise floor fails
Noise is common in a home studio, and every way of making the voice louder makes it worse. In one forum thread a narrator measured −75 dB of noise at −38 dB RMS, and a reply pointed out that raising that file by 18 dB raises the noise floor by 18 dB, which puts it out of spec.
- Fix the room and the distance first. In that thread, blankets over the booth and a closer, slightly off-axis microphone gave −80 dB of noise at −25 dB RMS.
- Use the RMS window. Loudness Normalization raises noise and voice by the same amount, so a target of −22 dB instead of −20 dB lowers the noise floor by 2 dB and still passes. The Support guide suggests aiming for noise quieter than about −65 dB.
- Then apply gentle noise reduction. In Audacity 4 it is Effect ▸ Noise removal and repair ▸ Noise reduction, and its defaults of 6 dB, sensitivity 6 and 6 bands are the Support guide’s starting point. Try 9 dB next. At 12 dB the voice can turn “wine-glassy or honky”, in the guide’s words, and fail review. Run Step 2 again afterwards, since removing noise lowers the average. Noise reduction in Audacity walks through the profile and the sliders.
Over-processing fails human review
After the automatic check, a person listens, and the Support guide warns that a voice beaten into passing the numbers comes back marked “Overprocessing”. In a forum critique of a narrator’s sample, an experienced helper said the theatrical leads and tails of each word were missing, the soft starts and ends that heavy noise processing cuts away.
The Support guide adds that mastering does not touch the performance. Harsh sibilance, a gritty voice and wet mouth noises come from the reading, and the chain leaves them as they are. Breaths and mouth clicks are an edit, not a mastering step; how to remove breaths and mouth clicks in Audacity covers them. Use as few corrections as the numbers allow, and compare the result with your raw backup at the same loudness. Truncating every pause to hide the noise fails the same way, because the numbers pass and the chapter sounds unnatural.
Common mistakes
- Making an MP3 from an MP3. Re-encoding lowers quality in unpredictable ways. Keep the work in WAV and encode the MP3 once, as the last step.
- Exporting at a variable bit rate. ACX asks for constant. The Audacity 4 default, Preset, is not constant.
- Reaching the RMS target with Amplify or Normalize. Both move peaks and average together, so the peaks pass −3 dB before the average reaches −20 dB. Loudness Normalization followed by the limiter separates the two.
Pro tips
- Restart the sentence after a mistake. Go back to the beginning of the sentence or section, not one word back, so the edit lands at a natural breath. In Audacity 4, Record ▸ Lead-in Recording (Shift+D) plays the seconds before the playhead and then records, so you come in on cue.
- Record at a sensible level. The Support guide asks for occasional peaks around −6 dB in a quiet room with no echo. How to record audio in Audacity covers the meter and the input level.
Example settings
| Step | Audacity 4, one chapter | Audacity 3, Audiobook Mastering macro |
|---|---|---|
| Low cut | Filter Curve EQ, Low rolloff for speech | Filter Curve EQ, Low rolloff for speech |
| Loudness | Loudness Normalization, RMS, −20 dB | Loudness Normalization, RMS, −20 dB |
| Peaks | Limiter in the Master section, Threshold and Output −3.5 dB | Limiter, Soft Limit at −3.5 dB, no make-up gain |
| Export | MP3 Files, Constant, 192 kbps, 44100 Hz | MP3, Constant, 192 kbps |
| Check | The exported MP3 in ACX Check | ACX Check on the project, then on the MP3 |
If the noise floor fails, lower the RMS target before you reach for more noise reduction.
Voice plugins and the Voice Extractor
The built-in effects are enough to pass the numbers. The plugins below help with what the numbers miss, in light doses. They install through MuseHub and appear under Add effect after the next start; like the Master limiter, they act on the export, so measure it as in Step 5.
Viator Voice Denoiser by Viator DSP is a one-knob noise reducer tuned for spoken voice. It runs live, so you hear the result as you turn the knob, with no noise profile to capture. Stop as soon as the hiss drops; a step further makes the voice hollow.
MuseFX is Muse Group’s free effects collection, and its de-esser is the tool for the harsh “s” the mastering chain leaves alone. A de-esser turns down only the sibilant frequencies, only when they spike, so the rest of the voice keeps its tone.
For a take recorded over background sound, such as a television in the next room, the online Voice Extractor on MuseHub keeps the voice and drops the rest. It is free and runs in the browser on MP3, WAV, FLAC or M4A files of up to 1 GB or 20 minutes. Use it as a last resort; re-recording the passage is safer, because the extracted voice still has to pass the human listener. The best podcasting plugins and the best vocal plugins cover the wider field, and the rest is in plugins on MuseHub.
FAQ
Next steps
How to use Audacity covers the rest of the editor, and how to record audio in Audacity gets the next chapter right at the source. Get Audacity 3, Audacity 4 and the voice plugins in one place — download MuseHub free.