The gate your files must pass
ACX audio requirements: the complete 2026 checklist
ACX accepts audiobook files that hit three measured targets — RMS between −23 and −18 dB, peaks below −3 dB, noise floor below −60 dB RMS — delivered as 192 kbps+ CBR MP3s at 44.1 kHz, with 1–5 seconds of room tone at each end of every file. Everything below is pulled from ACX's own published requirements and was re-verified on 2026-09-20.
Specs on this page were read directly from ACX's help documentation on 2026-09-20. ACX can revise its requirements at any time; if you're mastering a submission months from now, spend two minutes re-checking their current help pages before you export 40 finished files. When our page and ACX's page disagree, ACX wins.
The measured specs
Three numbers get measured on every file you upload. All three are checkable at home, free, before you submit — the method is in our measurement guide.
| Spec | Requirement | What it means in practice |
|---|---|---|
| RMS level | Between −23 dB and −18 dB | Average loudness of your narration. Too low and listeners crank the volume; too hot and it fatigues. Most chains hit it with normalization or gentle compression plus a final RMS check. |
| Peak level | Below −3 dB | No single moment may come within 3 dB of digital maximum. A limiter set at −3.2 dB as the last step in the chain makes this automatic. |
| Noise floor | Below −60 dB RMS | Your room-plus-electronics, measured where you aren't speaking, must sit at least 60 dB under full scale. The hardest of the three for home narrators — it gets its own deep dive. |
All three are checkable at home, free, before you submit.
File format and delivery rules
The format rules are pass/fail and entirely mechanical — no artistry, just settings at export time:
- 192 kbps or higher MP3, constant bit rate (CBR), 44.1 kHz. Variable bit rate exports are a common silent failure — check your encoder says CBR, not VBR or ABR.
- All mono or all stereo, consistently. Pick one for the whole book. Mono halves upload size and is the working standard for single-narrator audiobooks; whichever you choose, no mixing formats mid-book.
- No file longer than 120 minutes. Split monster chapters.
- One chapter or section per file, with the section header (“Chapter 1”) read aloud at the top.
- Filenames in standard US alphanumeric characters — no accents, no emoji, no cleverness.
Room tone: 1–5 seconds, both ends, every file
Each uploaded file must open and close with 1 to 5 seconds of room tone — the sound of your recording space with you in position and silent — and must not exceed 5 seconds of it. Two practical notes the spec implies but narrators discover the hard way. First, room tone means your room's tone, recorded in session, not inserted digital silence: a file that opens with mathematically dead silence and then lifts to an audible floor when you speak sounds edited, because it is. Second, keep a 30-second room tone recording from every session in your project folder; when edits create gaps, you patch with matching tone instead of silence. Tone that doesn't match its surroundings is one of the easiest tells for QA to hear.
- 192 kbps+ CBR MP3
- 44.1 kHz
- All mono or all stereo
The performance and content rules
Beyond the meters, ACX requires the recording itself to hold up:
- Free of extra sounds — the spec names plosives, microphone pops, mouse clicks, excessive mouth noise, and outtakes. Your floor can pass while a stray mouse click fails the file; we wrote up that exact trap.
- Consistent sound and formatting across the whole book — levels, tone, noise, spacing, and pronunciation shouldn't drift between chapters recorded weeks apart. Consistency is a measurement discipline: same gain, same distance, same room state, verified per session, which is exactly what locked-down gain staging buys you.
- Opening credits stating title, author, and narrator; closing credits in the “You have been listening to…” form.
- A retail sample of 5 minutes or less, ideally from the book's opening — this is the storefront preview, so make it the cleanest 5 minutes you own.
- Human narration. As of our 2026-09-20 check, ACX's requirements prohibit unauthorized text-to-speech, AI, or automated recordings. The gate you're mastering for is explicitly a gate for people.
The submission-day checklist
Run this against your finished files, in this order — it's sequenced so the expensive checks come last:
- Every file 192 kbps+ CBR MP3, 44.1 kHz, consistent mono/stereo, under 120 minutes.
- 1–5 seconds of genuine room tone at head and tail of every file.
- Measured on every file: RMS −23 to −18 dB, peaks under −3 dB, floor under −60 dB RMS.
- Spot-listen the first and last minute of every file at full attention, headphones on, for clicks, pops, and edit seams.
- Credits present, section headers read, retail sample chosen and re-listened.
If step 3 is where you fail — and for home narrators it usually is — the fix isn't at submission time. Walk the fix order from the top; it exists precisely so this checklist becomes boring.
Frequently asked
- Are these requirements the same for every audiobook retailer?
- No — they're ACX's. But they're the strictest widely-published spec in consumer audiobooks, which makes them the sensible home target: files that clear ACX clear almost anything a direct client asks for.
- Does ACX check every file with software or by ear?
- Both realities apply to you: automated measurement catches spec numbers, and human quality review catches what meters miss. Master for the meter and the listener — the spec's “free of extra sounds” language is enforced by ears.
- What's the most common home-studio failure?
- The noise floor, by a wide margin — it's the only spec your room votes on. RMS and peaks are set in mastering; the floor is set by whatever your space and gain staging are doing at record time.
- Can I fix a failing floor by just turning the whole file down?
- No — lowering everything drops your RMS out of its own required range. The floor spec is effectively a ratio demand: voice loud enough, room quiet enough, at the same time. That's why it's the honest measure of a recording space.