Audible’s ACX audio requirements, explained

What the requirements are for

ACX (Audiobook Creation Exchange) is the service through which audiobooks are submitted for sale on Audible, Amazon and Apple Books. Every file is reviewed against its Audio Submission Requirements before the book goes on sale. Software measures some of them. ACX’s reviewers judge consistency and extra sounds by ear.

Loudness: -23 dB to -18 dB RMS

RMS (root mean square) is the average level of the whole file, not its loudest moment. ACX requires every file to measure between -23 dB and -18 dB RMS.

The range exists so that a listener does not have to change the volume from one chapter to the next, or from one book to the next. A file below -23 dB is too quiet for a car or a train. A file above -18 dB got there by hard compression.

How to meet it: record at a healthy level, with peaks around -12 dB to -6 dB while you read, and raise or lower the whole chapter to the target afterwards. A small amount of compression evens out the louder and quieter passages. Raising the level raises the noise floor with it, so measure the two together.

Peaks: below -3 dB

The peak is the single loudest sample in the file. ACX requires it to stay below -3 dB, which leaves room for the conversion to the formats Audible streams and sells. A peak that touches 0 dB can distort in that conversion.

How to meet it: avoid clipping while recording, since a clipped peak cannot be repaired cleanly. A limiter set to -3 dB, or slightly below, holds the peaks under the line after the level has been raised to the RMS target.

Noise floor: below -60 dB RMS

The level of the recording when nobody is speaking: the room, the computer fan, the street and the microphone’s own hiss. ACX requires it below -60 dB RMS, measured in the pauses between words.

This requirement is the hardest one to meet in a home room, and it is decided mostly before recording. A quiet room, a microphone close to the mouth and a microphone that picks up little of the room do more than any noise reduction afterwards. Heavy noise reduction lowers the measured floor but can leave the voice thin or watery, which ACX’s reviewers also hear.

How to meet it: turn off fans and appliances, record in a treated space and keep the microphone a hand’s width from the mouth. Use gentle noise reduction only where it is needed. The room treatment guide covers what works at each budget, and the microphone guide lists microphones that reject an untreated room.

File format: MP3, 192 kbps or higher, constant bit rate, 44.1 kHz

256 kbps and 320 kbps are accepted too. ACX requires a constant bit rate so that its own encoding process runs without errors; a variable bit rate (VBR) file is not accepted.

How to meet it: record and edit at any quality you like, and export the finished chapters as 44.1 kHz CBR MP3. A recording made at 48 kHz is converted to 44.1 kHz on export.

Mono or stereo, the same for every file

One voice is recorded as a single channel; a two-channel copy of it doubles the size and adds nothing.

Room tone at the start and end of each file

The sound of the recording space with nobody speaking. ACX recommends between 1 and 5 seconds of it at the beginning and the end of each file, and states that room tone spacing must not exceed 5 seconds.

The pause is room tone and not digital silence, because a file that drops to nothing sounds as if the audio has cut out. The head gives the listener a moment before the first word; the tail separates one chapter from the next.

How to meet it: record a few seconds of the room at the start of each session, and put that sound, not silence, at the head and tail of each chapter.

File length: no longer than 120 minutes

A chapter or section longer than that is split into two files, and the second carries a secondary header so the listener knows the chapter continues, for example “Chapter 12, continued”.

One section per file, and the file names

Each chapter or section is uploaded on its own, labeled with its title, such as “Prologue” or “Chapter 1”. File names use standard US letters and numbers, nothing else.

Opening and closing credits

The opening set states the audiobook’s title, its author or authors and its narrator or narrators, and must match the book’s metadata and cover art. The closing credits signal the end of the book. ACX recommends “You have been listening to” followed by the title, the authors and the narrators, ending with “The End.”

The opening and closing credits are separate files from the chapters. Music is allowed in the credits if it does not cover the narration.

The retail sample

The excerpt a customer can play before buying. It must be five minutes or less, it must come from the audiobook itself, and it must not contain explicit material, such as violent or pornographic passages. ACX asks for the first five minutes of the audiobook as the sample.

Consistency and extra sounds

These two requirements are judged by ear. The sound of the book must be consistent from file to file: in level, tone, background noise, the spacing between lines and the pronunciation of names. And there must be no extra sounds: plosives (popped P and B sounds), microphone pops, mouse clicks, excessive mouth noise and outtakes are all removed.

How to meet it: record every session in the same place, with the same microphone at the same distance and the same settings, and listen to the whole book before submitting it. A change of room or microphone halfway through a book is audible.

Human narration

ACX requires the submitted audiobook to be narrated by a human unless ACX has authorized otherwise. Unauthorized text-to-speech, AI-generated and automated recordings are not accepted. Other stores have their own rules for AI narration, and those rules are stated on each store’s own pages.

How StellAurator checks a book before export

In StellAurator, choose the output Audiobook master for Audible’s ACX and press the five steps in order: Diagnose, Plan, Apply, Check, Export. Each chapter is measured against ACX’s numbers for loudness, peaks, noise floor and file format, before and after processing, with a pass or fail on each.

After each take, the clicks, plosives, breaths, hum and background noise are marked under it with a suggested fix for each. You hear each fix before you keep it, and a kept fix is undone with one press. The export writes one file per chapter in the format ACX requires.

ACX decides what it accepts. Its reviewers also listen to the performance and check the rights to the text.