Audiobook

ACX audio requirements: how to prepare and master an audiobook for Audible

The exact RMS, peak and noise-floor specs, the file structure ACX expects, and a mastering workflow that clears ACX's automated technical check on the first upload.

By Hanna Eng·Audio engineer, Abbey Road Institute Paris

Updated 1 June 202610 min read

To pass ACX, submit a mono (or stereo) MP3 at 192 kbps CBR (minimum) and 44.1 kHz with an RMS level between -23 dB and -18 dB, a peak no higher than -3 dB, and a noise floor below -60 dB. Each section is its own file under 120 minutes, with 0.5 to 1 second of room tone at the head and 1 to 5 seconds at the tail.

ACX rejects audiobooks before a human ever listens. An automated check measures your loudness, peaks, noise floor, format and file structure, and a single value out of range bounces the whole upload. This guide lays out every requirement with its exact value, then shows how to master to it cleanly, so your submission clears the automated technical check on the first upload and avoids the most common technical rejections. Meeting the numbers is necessary but not sufficient: a separate human quality review still follows.

ACX audio specs at a glance

SpecRequired value
RMS (average level)-23 dB to -18 dB
Peak-3 dB maximum
Noise floor-60 dB RMS or lower
FormatMP3, constant bit rate (CBR)
Bit rate192 kbps CBR (minimum)
Sample rate44.1 kHz
ChannelsMono recommended, stereo accepted
File lengthUnder 120 minutes each
Room tone0.5-1 s head, 1-5 s tail
AI / synthetic narrationNot allowed (human narration required)
Filename charactersLetters, numbers, dashes and underscores only

Source: ACX, Audio Submission Requirements (official); ACX, Mastering Audiobooks with Alex the Audio Scientist (official)

What is ACX?

ACX (Audiobook Creation Exchange) is Amazon's platform for producing and distributing audiobooks to Audible, Amazon and Apple Books. Every file you upload passes an automated technical review covering volume, peaks, noise floor, format and structure. Any file outside the spec is rejected before the human quality check even begins, so hitting the numbers is step one.

The ACX technical specs at a glance

A compliant file is a mono (or stereo) MP3 at 192 kbps constant bit rate (minimum) and 44.1 kHz, mono recommended and stereo accepted, with an RMS level between -23 dB and -18 dB, a peak no higher than -3 dB, and a noise floor below -60 dB. It opens with 0.5 to 1 second of room tone and ends with 1 to 5 seconds.

  • RMS (average level): between -23 dB and -18 dB.
  • Peak: maximum -3 dB.
  • Noise floor: -60 dB RMS or lower.
  • Format: MP3, constant bit rate (CBR).
  • Bit rate: 192 kbps CBR (minimum).
  • Sample rate: 44.1 kHz.
  • Channels: mono recommended, stereo accepted, but the whole book must be consistent (never mix mono and stereo files).
  • Each file under 120 minutes.

RMS, peak and noise floor: what they measure

RMS is the average level of your narration over time, the closest meter to perceived volume. Peak is the single loudest sample. Noise floor is the level of the quietest part, your room tone, where no one is speaking.

A useful nuance: ACX expresses everything in dBFS, and the RMS dB the spec refers to is dBFS RMS, not LUFS. Audiobooks predate the LUFS standard used for streaming music, which is why the yardstick is different. In practice you aim for around -20 dBFS RMS, comfortably inside the -23 to -18 window.

Structuring and naming your files

ACX requires one file per section: opening credits, each chapter, and closing credits. You cannot upload the whole book as a single file, and each file must run under 120 minutes. Number your filenames (001, 002, and so on) so the playback order is unambiguous.

  • Opening credits (separate file): title, author and narrator, spoken aloud.
  • Closing credits (separate file): the closing line (for example "The End").
  • Room tone: 0.5 to 1 second at the head, 1 to 5 seconds at the tail of every file.
  • Retail sample: 1 to 5 minutes, starting with narration (not credits or music), no explicit content.
  • Cover art: square, 2400 x 2400 px minimum.

Mastering to ACX, step by step

ACX's own recommended chain is simple: clean unwanted frequencies with EQ, tame peaks with a limiter to create headroom, then raise overall gain until the file sits inside the -23 to -18 dB RMS window. The practical target is around -20 dB RMS, a real peak under -3 dB, and a noise floor below -60 dB.

Set loudness last. The most common mistake is normalizing a noisy recording, which simply amplifies the problems. Clean first, then level.

  • Reduce the noise floor below -60 dB with gentle passes (iZotope RX, 6 to 8 dB at a time rather than one aggressive pass).
  • Reach the RMS window without crushing the natural dynamics of the voice.
  • Measure the noise floor by selecting 3 to 5 seconds of silence and reading its RMS, then check overall RMS and peak.

Tools to measure and fix

A professional station like Pro Tools handles recording and editing. For the MP3, the reliable workflow is to bounce a 24-bit WAV master and encode to MP3 with CBR explicitly set to 192 kbps and 44.1 kHz, in the DAW or a dedicated encoder, then verify the final MP3, rather than assuming a default DAW MP3 bounce meets the CBR spec. iZotope RX profiles and reduces the room noise and reads the file statistics (noise floor, RMS, peak). Audacity with the free ACX Check plug-in is an entry-level option for verifying the three numbers before upload.

The errors that get an audiobook rejected

Most ACX rejections come down to a handful of measurable faults. Fixing them is mechanical once you know what the check is looking for.

  • RMS outside the -23 to -18 dB window (raise or lower overall gain).
  • Peak above -3 dB (apply a limiter with a -3 dB ceiling).
  • Noise floor above -60 dB (denoise the room tone in RX).
  • Missing or excessive room tone (trim to 0.5-1 s head, 1-5 s tail).
  • Wrong bit rate or sample rate (export 192 kbps CBR, 44.1 kHz).
  • Audible artifacts left in the track (mouth clicks, breaths, hiss).

Submitting your audiobook on ACX

Before any upload, you must link the print or Kindle edition of your book on Amazon. You then create an ACX project, upload each section file, add the square cover (2400 px minimum) and the retail sample. ACX does not publish a fixed review window, and turnaround times vary, so plan for the review to take some time rather than counting on a specific number of days.

ACX royalty tiers: exclusive vs non-exclusive

Royalties are a platform term, separate from the audio specs, but creators preparing a title ask about them constantly. ACX has historically paid a higher rate for exclusive distribution (sold through Audible, Amazon and Apple Books only) than for non-exclusive distribution, where you can also sell the title elsewhere. Platform terms change over time, so treat the figures below as the established baseline and confirm the rate tied to your title in the ACX dashboard.

The figures below are the neutral platform percentages, not a recommendation. The right tier depends on your own distribution plans, and the exact rate tied to a title appears in the ACX dashboard.

DistributionRoyalty
Exclusive (Audible/Amazon/Apple Books only)40%
Non-exclusive (sell elsewhere too)25%

Source: ACX royalty options (acx.com); confirm the current rate in your ACX dashboard

Frequently asked questions

What are the ACX audio requirements?

A mono (or stereo) MP3 at 192 kbps CBR (minimum) and 44.1 kHz, with RMS between -23 and -18 dB, a peak no higher than -3 dB, and a noise floor below -60 dB. Mono is recommended and stereo is accepted. Each section is its own file under 120 minutes, with 0.5 to 1 second of room tone at the head and 1 to 5 seconds at the tail.

What RMS level does ACX require?

Between -23 dB and -18 dB RMS, measured in dBFS. Aiming for around -20 dB RMS keeps you safely in the window with margin on both sides.

What is the ACX noise floor and how do I measure it?

The noise floor is the level of your silent room tone, and ACX requires it to be -60 dB RMS or lower. Measure it by selecting 3 to 5 seconds where no one speaks and reading the RMS of that selection.

Does ACX require mono or stereo?

Mono is recommended for spoken-word audiobooks, but stereo is accepted. The key rule is consistency: the entire book must be one or the other, never a mix of mono and stereo files.

Why was my audiobook rejected by ACX?

Usually one measurable value is out of range: RMS outside -23 to -18 dB, a peak above -3 dB, a noise floor above -60 dB, missing or excessive room tone, the wrong bit rate or sample rate, or audible noise left in the track.

Does ACX require 16-bit or 24-bit audio?

ACX does not state a bit depth in its submission requirements. The submitted file is an MP3 at 192 kbps CBR and 44.1 kHz, and MP3 is a compressed format where bit depth is not specified. Bit depth only matters at the recording stage: capture your WAV master in 24-bit for editing headroom, then export to the required 192 kbps MP3. Nothing on ACX's pages asks for a 16-bit minimum on the delivered file.

Is the -3 dB peak a hard ceiling, and how do I stay under it after MP3 export?

Yes. ACX wants peaks under -3 dB to prevent distortion, so treat -3 dB as a ceiling you stay below, not a target you hit. Set a limiter with a -3 dB ceiling before you encode, and leave a little margin (aim for a sample peak around -3.5 to -4 dB before encoding) because MP3 encoding can nudge the peak level slightly higher. ACX measures sample peak (dBFS) here, not true peak, so always re-measure the peak on the final MP3, not just on the WAV, before you upload.

What royalty does ACX pay, exclusive vs non-exclusive?

ACX has historically paid 40% on exclusive distribution and 25% on non-exclusive distribution. These are neutral platform terms, not advice on which to choose, and platform rates can change over time, so check your own ACX dashboard for the rate tied to your title.

Can I use an AI voice to narrate an ACX audiobook?

No. ACX requires human narration and does not accept text-to-speech, AI or other automated recordings for general submissions. The one exception is a limited, US-only beta in which invited narrators can create an AI replica of their own voice; those titles are labelled in the narrator field and the narrator controls their use title by title. Outside that beta, an AI-narrated file will be rejected.

Can I use accents or special characters in my ACX filenames?

No. ACX advises sticking to alphanumeric characters, dashes and underscores in filenames; other characters can cause upload issues. Avoid accented letters (é, à, ç), spaces, slashes and symbols. Number each section first so the order is unambiguous, for example 00_Opening-Credits, 01_Chapter-01, 02_Chapter-02.

Sources and references

Have a project that needs this done right?

If your mix has to pass a platform spec, let's talk about the deliverables and the timeline.

Start a project