ACX audio requirements, explained

ACX is Audible’s marketplace for audiobook production, and every file you upload goes through a technical check before a person listens to it. The check is strict but not mysterious. There are a handful of measurements, each with a clear limit, plus some rules about how the book is split into files. Here they all are, as ACX publishes them, with what each one means in practice.

The requirements at a glance

From ACX’s Audio Submission Requirements. ACX updates its rules from time to time, so check that page before a big delivery.

Loudness: RMS between −23 and −18 dB

RMS (root mean square) is an average of how loud the whole file is, not its loudest moments. ACX wants every chapter in a fairly narrow window so listeners don’t have to reach for the volume between chapters.

Aim for the middle of the window, around −20 dB. That leaves room on both sides: a chapter with a lot of quiet, intimate reading measures lower than one full of shouting, and you don’t want either to fall out of range. What matters is that every file lands inside it.

You rarely get there by recording louder. You get there in mastering, by evening out the reading (compression or a leveler) and then raising the whole file to the target.

Peaks: no higher than −3 dB

The peak is the single loudest instant in the file. ACX wants it at least 3 dB below the maximum a digital file can hold. The reason is that MP3 encoding and playback devices can push peaks a little higher than they were in your edit, and that headroom stops them from distorting.

The standard way to guarantee it is a limiter as the very last step of mastering, set a little below the limit (−3.5 dB is common), so the encoded MP3 still measures under −3 dB.

Noise floor: no higher than −60 dB

The noise floor is how loud your recording is when you’re not speaking: room hum, computer fans, traffic, the hiss of the microphone and preamp. ACX measures it in the pauses. −60 dB is quiet; most rejections for noise come from fans, air conditioning or a microphone gain turned up too far.

• Fix it at the source first. Turn off fans and air conditioning while recording, move the computer away from the microphone, and treat the room with soft furnishings.
• Don’t record too quietly. If you have to boost a quiet recording a lot in mastering, the noise comes up with it.
• Use noise reduction gently. Heavy noise reduction makes a voice sound watery and metallic, which a listener notices far more than a little room sound.
• Don’t chase digital silence. Cutting pauses to complete silence makes them sound like dropouts. Pauses should sound like your room, just quiet. That’s what room tone is for (see below).

File format: 192 kbps CBR MP3 at 44.1 kHz

Upload MP3s encoded at a constant bit rate (CBR) of 192 kbps or higher, at a 44.1 kHz sample rate. Variable bit rate (VBR) files are rejected. Higher bit rates such as 256 or 320 kbps are accepted, but listeners won’t hear a difference for spoken word.

Record and edit in WAV (24-bit is ideal) and only make the MP3 at the very end. Every time a file is re-encoded as MP3 it loses a little quality.

Mono or stereo, but never both

A single narrator is naturally mono, and ACX accepts mono or stereo. What it won’t accept is a book where some files are mono and some are stereo. If you export chapters from different programs or computers, check that they all came out the same way.

Room tone at the start and end

Each file should begin and end with 1 to 5 seconds of room tone, the recorded sound of your room with nobody speaking. It gives every chapter a clean, natural start and finish, and it’s what a listener hears between chapters. Pauses inside a chapter shouldn’t run longer than about 5 seconds either. Our room tone guide covers how to record it.

How the book is split into files

• One chapter or section per file, each starting with its title read aloud: “Prologue”, “Chapter One”.
• No file longer than 120 minutes. Split a very long chapter into parts.
• Opening credits as their own file: the title, the author or authors, and the narrator.
• Closing credits as their own file, confirming the end of the book by restating the same details.
• A retail sample of 5 minutes or less, which listeners hear before they buy. The opening of the book is usually best, and it must not contain explicit material.

Why chapters get rejected

1. Noise floor too high, usually from a fan, air conditioning, or a quiet recording boosted a lot.
2. Peaks over −3 dB because there was no limiter, or it was set right at −3 and the MP3 encoding pushed a few peaks over.
3. RMS out of range on quiet or very dynamic chapters.
4. Missing or inconsistent room tone, or pauses cut to dead silence.
5. Mixed mono and stereo files, or files that aren’t 44.1 kHz CBR MP3s.
6. Missing credits, or credits included inside the first or last chapter instead of as separate files.

Passing the technical check is necessary, not sufficient. ACX’s reviewers also listen for things a meter can’t catch: mouth clicks, distracting breaths, mispronunciations, inconsistent sound between sessions and missing words.