ACX audio requirements, explained

The short list

Beyond the numbers, ACX asks for one chapter per file with its spoken header, opening and closing credits as their own files, a retail sample of five minutes or less, consistent sound from file to file, no outtakes or mouth noise, and a human narrator.

Loudness: RMS between −23 dB and −18 dB

RMS is the average loudness of the whole file, pauses included. Too quiet and listeners turn the volume up into the hiss; too loud and the voice is squashed. ACX's window is five decibels wide, so aim for the middle, about −20 dB, and every chapter lands inside it even when one is read a little softer.

The usual way there is gain, then gentle compression, then a limiter. The trap is measuring a chapter that ends in a long silence or begins with a loud knock: both move the average. Measure the finished MP3, not the project.

Peaks: no higher than −3 dB

The peak is the single loudest sample in the file. ACX wants it at −3 dB or below, which leaves room for the MP3 encoder: encoding can push a peak up by a few tenths, so a file limited to exactly −3.0 dB can come back at −2.8. Set the limiter a little under, at −3.5 dB, and check the MP3 itself.

Noise floor: no higher than −60 dB RMS

The noise floor is how loud your room is when you are not speaking: the fridge, the fan, the computer, the street. ACX wants it at −60 dB or quieter. This is the requirement home recordings fail most often, and it is fixed mostly in the room: soft furnishings, a closet full of clothes, the computer's fan away from the microphone, the heating off for the take.

Noise reduction helps with a steady hum or hiss, but pushed too far it makes the voice sound underwater, which ACX's reviewers reject too. Turning the gain down does not lower the noise floor relative to your voice: once the file is brought up to −20 dB RMS, the noise comes up with it.

Room tone: 1 to 5 seconds at the start and end

Each file should begin and end with the quiet of your room, recorded, not inserted silence: ACX recommends 1 to 5 seconds and says spacing must not go over 5 seconds. The same rule is why a pause should never drop to dead digital silence: a listener hears it as the audio cutting out. Record half a minute of your room at the start of each session and use it to fill.

The file: 192 kbps CBR MP3 at 44.1 kHz

Every file must be an MP3 at 192 kbps or higher, constant bit rate (not variable), at 44.1 kHz. 256 and 320 kbps are allowed. Mono or stereo are both accepted, but every file in the book must be the same; a voice recorded on one microphone is mono.

One chapter per file, 120 minutes at most

Each file holds one chapter or section, and the narrator says its header ("Chapter Three"). A file may be 120 minutes long at most; split a longer section and give the second part a header for continuity. Use plain letters and numbers in file names.

Opening credits, closing credits and the retail sample

The opening credits are their own file and say the title, the author and the narrator, matching the title's details and cover. The closing credits are their own file too, and end the book clearly: "You have been listening to…", the title, author and narrator again, then "The End". The retail sample is five minutes or less, usually the first five minutes, with nothing explicit in it.

Consistency, clean sound and a human narrator

ACX asks that the sound stay consistent from file to file in level, tone, noise and spacing; that files be free of plosives, microphone pops, mouse clicks, mouth noise and outtakes; and that the narrator be a person. Unauthorised text-to-speech or AI narration is not allowed.

How to check before you upload

ACX offers its own audio analysis tool in its Audio Lab, and Audacity has an ACX Check plugin. Check the MP3s you will upload, after export, not the recording project. Instant ACX Ready measures each MP3 it makes for loudness, peaks, noise floor, room tone at both ends and length, and shows any number outside ACX's window in red.