ACX audio requirements, explained

ACX checks every file against a short list of technical rules before a human ever listens. Here is each rule, what it means in practice, and how to meet it.

Checked against ACX's audio submission requirements in October 2026. ACX can change its rules; their page is the final word.

The numbers at a glance

Loudness: RMS between -23 and -18 dB

RMS is a measure of average loudness across the file. Most home recordings come in too quiet, somewhere around -30 dB, because people set their microphone gain low to avoid clipping. That's the right instinct; the fix is to raise the level afterwards with gain and a limiter, not to shout or crank the mic.

Aim for the middle of the range, around -20 dB. That leaves room for the measurement to vary a little between chapters and still pass.

Peaks: no higher than -3 dB

Raising the average level also raises the loudest syllables. A limiter holds those peaks down while the rest comes up. Watch out for MP3 encoding: it can push peaks up slightly, so a file limited to exactly -3.0 dB before encoding may measure -2.8 dB afterwards and fail. Limit to around -3.5 dB, and measure the finished MP3, not the WAV.

Clipping at the microphone (the input hitting 0 dB) is a different problem. It's baked into the recording as distortion and no later processing can remove it. Re-record that line at a lower gain.

Noise floor: no higher than -60 dB

This is the rule home narrators fail most often. The noise floor is measured in the quietest stretches of your file, so it's really a measure of your room: computer fans, fridges, air conditioning, street noise and electrical hum.

The catch is that it interacts with loudness. If you record quietly and then add 12 dB of gain to reach -20 dB RMS, your room noise goes up by 12 dB too. A room that measures -70 dB raw can end up at -58 dB after gain and fail. Two fixes, in order of preference:

1. Make the room quieter: switch off fans and appliances, record away from windows, and hang blankets or use a closet full of clothes to cut reflections.
2. Record a bit louder (closer to the mic, or more gain) so less gain is needed afterwards.

Noise reduction can rescue a borderline file, but heavy noise reduction makes voices sound watery, and ACX listeners notice. Gentle is better.

Format: 192 kbps CBR MP3 at 44.1 kHz

Export MP3 at a constant bit rate of 192 kbps (256 or 320 kbps are also accepted) and a sample rate of 44.1 kHz. Many tools default to variable bit rate or 48 kHz, so check the export settings. Keep every file in the book the same: all mono or all stereo.

Room tone at the start and end

Each file should open and close with 1 to 5 seconds of your room's natural background sound, not digital silence. Digital silence sounds like the audio cut out and makes the noise floor jump between words and pauses. Capture a few seconds of your room while sitting still, and use it to pad each chapter.

Structure: one chapter per file, plus credits

• One chapter or section per file, uploaded individually, each starting with its section header read aloud ("Chapter 1", "Prologue").
• Opening credits in their own file: the title, author(s) and narrator(s), matching your metadata and cover.
• Closing credits in their own file, ending with "The End."
• A retail sample of five minutes or less, taken from the audiobook itself, with no explicit material.
• File names with plain letters and numbers only.

What ACX checks by ear

Passing the numbers gets you past the automated check. ACX's reviewers also listen for consistency across the book (levels, tone, noise, pacing, pronunciation) and for sounds that shouldn't be there: plosives and mic pops, mouth clicks, mouse clicks, page turns and outtakes. No measurement tool can catch all of those; listen to every chapter on headphones before you upload.