Updated March 2026 · 12 min read · ACX Submission Guide · By Daniel Shilansky, Founder, TomeVox

ACX Technical Requirements for Audiobooks: Full 2026 Guide

ACX — Amazon's audiobook distribution marketplace — requires audio files to meet five technical specs: RMS loudness between -23 and -18 dBFS, peak level at or below -3 dBFS, noise floor at or below -60 dBFS, 44.1 kHz sample rate, and 192 kbps constant bit rate (CBR) MP3 format. Any file outside these ranges is rejected automatically before a human reviewer sees it.

Getting rejected by ACX after weeks of production work is one of the most frustrating experiences an audiobook creator can face. The ACX quality check is thorough, automated, and unforgiving — a noise floor that's 2 dB too loud or a peak level that clips at -2.9 dB will get your submission bounced, no matter how good the narration sounds to human ears.

ACX audiobook submissions must meet specific technical specifications for sample rate, loudness, noise floor, peak level, and file structure. All level measurements use dBFS — decibels relative to full scale, where 0 dBFS is the absolute ceiling of digital audio and all compliant values are negative numbers. The sections below explain what each measurement means, walk through the most common rejection reasons, and show how to verify files before submission. Authors using TomeVox for audiobook production can skip most of this — TomeVox handles ACX compliance automatically.

What are the ACX technical requirements?

Every audio file submitted to ACX must pass five checks: RMS loudness between -23 and -18 dBFS, peak level at or below -3 dBFS, noise floor at or below -60 dBFS, 44.1 kHz sample rate, and 192 kbps CBR MP3. The full specification is published on ACX's audio submission requirements page. The table below shows all required values:

Key finding: The noise floor requirement (-60 dBFS or lower) is the most difficult specification to meet in practice and the most common reason ACX rejects human-recorded audiobooks. AI-generated audio — which has no room or microphone noise — meets this specification by default.

What does each ACX specification actually mean?

Sample Rate: 44.1 kHz Is Non-Negotiable

44.1 kHz is the only sample rate ACX accepts — submitting a file at 48 kHz will trigger an automatic rejection or a forced downconversion that introduces audio artifacts. The sample rate is the number of audio samples captured per second; CD-quality audio uses 44,100 per second, while video and broadcast audio commonly uses 48,000 Hz. Always set your DAW project to 44.1 kHz before you begin recording — changing the sample rate after the fact introduces degradation.

Noise Floor: -60 dBRMS Is the Hardest Spec to Hit

The noise floor is a measurement of how quiet your "silence" actually is. In practice, all rooms have ambient noise: HVAC hum, electrical interference from computers, traffic, building creaks. ACX measures the RMS level of a quiet passage in your audio and compares it against the -60 dBRMS threshold.

Most home recording environments without acoustic treatment fall between -50 and -55 dBRMS — which will fail ACX review. In our experience working with authors through the ACX submission process, noise floor violations are consistently the most common rejection reason. Meeting -60 dBRMS typically requires either a properly treated recording space, noise reduction software (iZotope RX is the industry standard), or both.

Peak Level and RMS Loudness: Two Different Measurements

Peak level and RMS loudness are often confused but measure different things. Peak level measures the loudest single sample in your audio — the instantaneous ceiling. ACX requires no sample exceed -3 dBFS. A peak limiter set to -3 dB handles this automatically.

RMS loudness (sometimes expressed as LUFS — Loudness Units relative to Full Scale) measures the average perceived loudness over time. ACX requires your speech audio to average between -23 and -18 dBRMS. Too quiet and the listener has to crank their volume; too loud and it clips against ACX's ceiling. Most professional narrators target -20 dBRMS as a comfortable midpoint.

Room Tone: Opening and Closing Buffers

Room tone is a brief period of ambient silence at the start and end of each audio file. It serves two purposes: it gives ACX's automated QC system a clean sample to measure your noise floor against, and it provides a natural buffer so chapter transitions don't cut abruptly. Opening room tone should be 0.5 to 1 second. Closing room tone should be 1 to 5 seconds — enough to feel like a natural pause, not so long it feels like the track died.

How should you structure ACX chapter files?

ACX requires your audiobook to be submitted as individual audio files — one per chapter (or one per major section). This is different from a single continuous audio file. The structure must include:

• Opening Credits file (your name, title, copyright statement, publisher)
• One file per chapter, numbered sequentially
• Closing Credits file (narrator credit, publisher info)
• Retail Audio Sample (first 5 minutes of the book, this is auto-generated by ACX from your uploaded files)

File naming should be systematic. ACX does not enforce a specific naming convention, but clean naming prevents upload order errors. A reliable convention is:

00_opening-credits.mp3   01_chapter-01.mp3   02_chapter-02.mp3   99_closing-credits.mp3

Each chapter file should contain only one chapter. Do not concatenate multiple chapters into one file. ACX's system needs individual files to create proper chapter navigation in the Audible app.

Why does ACX reject audiobook submissions?

ACX publishes a list of common QC failure reasons, and the same issues come up repeatedly in audiobook production communities. The most frequent ACX rejection reasons are noise floor violations, peak level clipping, wrong sample rate, and missing room tone:

• Noise floor above -60 dBRMS — Home recording environments with untreated room acoustics, HVAC noise, or computer fan noise. By far the most common rejection.
• Peak levels exceeding -3 dBFS — Usually caused by loud consonants (plosives, sibilance) that weren't caught during editing, or music/effects tracks that weren't properly limited.
• RMS loudness outside the -23 to -18 dBRMS range — Too quiet is as bad as too loud. Narrators who record at conservative levels and don't normalize frequently get rejected for being too soft.
• Wrong sample rate — Submitting at 48 kHz instead of 44.1 kHz. This happens when narrators use video-oriented DAW templates.
• Missing room tone — Files that begin immediately with speech, or end immediately after the last word, without the required buffer.
• Inconsistent audio between files — Different noise floors, loudness levels, or EQ profiles between chapters, suggesting different recording sessions or environments.
• Background music or effects with no narration — Intro music, chapter-break stingers, or ambient soundscapes that are not approved by ACX guidelines. ACX allows music only under specific conditions.
• Audible edits — Clicks, pops, or abrupt cuts from sloppy editing that weren't caught in QC.
• Wrong file format or bit rate — Submitting Variable Bit Rate (VBR) MP3 instead of Constant Bit Rate (CBR), or submitting at 128 kbps instead of 192 kbps minimum.

How do you measure audio before submitting to ACX?

Before uploading a single file to ACX, run a self-QC pass on every chapter. The three critical measurements to verify are noise floor (must be -60 dBRMS or lower), peak level (-3 dBFS or lower), and integrated RMS loudness (between -23 and -18 dBRMS):

Measuring noise floor: In most DAWs (Audacity, Adobe Audition, Logic, Reaper), select a 3–5 second silent passage and run a statistics analysis. You're looking for the RMS level of that silent section. It must read -60 dBRMS or lower. In iZotope RX, the Meter module displays noise floor in real time.

Measuring peak levels: Use your DAW's built-in peak meter, or export to a loudness measurement tool like the free Youlean Loudness Meter. Look at the True Peak (TP) reading. It must be -3 dBTP or lower.

Measuring RMS loudness: Youlean Loudness Meter, the free LUFS Meter VST, or Adobe Audition's built-in Match Loudness panel will give you an integrated LUFS reading. Target -20 LUFS as a practical midpoint within ACX's -23 to -18 range.

How do you submit an audiobook to ACX step by step?

1. Create or claim your title on ACX Log in at acx.com, click "Add Your Title," and either link your Amazon KDP title or enter your book's metadata manually. You'll need the title, subtitle, author name, and ISBN if you have one. If you're self-publishing, you can create an ACX title without an ISBN.
2. Choose distribution and royalty type ACX offers exclusive distribution (Audible + Amazon + Apple Books) or non-exclusive. Exclusive pays 40% royalty; non-exclusive pays 25%. If you plan to distribute to Spotify Audiobooks, Kobo, or other platforms, choose non-exclusive now — you cannot change this after going live.
3. Prepare and verify all audio files Run your full self-QC pass on every file. Check noise floor, peak, RMS loudness, sample rate, bit rate, and room tone buffers. Fix any file that fails before uploading. One failed chapter will get the entire submission rejected.
4. Upload files in chapter order In your ACX dashboard, navigate to your title and click "Upload Your Audiobook." Upload files one at a time in chapter order — Opening Credits first, then Chapter 1, Chapter 2, and so on through Closing Credits. ACX processes each file as it uploads.
5. Preview each chapter after upload ACX provides an in-browser player for each uploaded file. Listen to the beginning and end of each chapter to confirm room tone is present, the audio loads cleanly, and there are no obvious artifacts from the upload process.
6. Submit for ACX Quality Check Once all files are uploaded and you're satisfied with the previews, click "Submit for Review." ACX's automated QC system runs first (usually within 24–48 hours), followed by a human review. The full process takes 7–14 business days.
7. Respond to QC feedback if rejected If your submission is rejected, ACX will email you with a specific reason code. Address only the flagged issues, re-export the affected files, and resubmit. You do not need to re-upload all files — only the chapters that failed.

Does TomeVox handle ACX compliance automatically?

TomeVox audiobook output is engineered to meet ACX specifications automatically — including 44.1 kHz sample rate, 192 kbps CBR, peak limiting at -3 dBFS, loudness normalization to -20 LUFS, and per-chapter file structure with room tone buffers. The full compliance breakdown:

The one thing TomeVox does not generate for you is the Opening Credits and Closing Credits audio files — those require your specific copyright statement, author name pronunciation, and narrator credit. You can record those yourself (they're typically 30–60 seconds each) and add them to the upload. Every other technical requirement is met by the TomeVox output.

For a complete walkthrough of the production process — from manuscript formatting to final distribution — see the TomeVox AI audiobook production guide.

The hardest part of ACX for human narrators — the noise floor — is simply not a problem for AI-generated audio. There is no room, no microphone, no HVAC system. The technical floor is clean by default.

Does ACX accept AI-narrated audiobooks?

Important: ACX's standard submission portal is designed for human-narrated titles. To publish an AI-narrated audiobook on ACX, you need to contact ACX support directly — do not attempt to submit through the standard flow without doing so first. The technical specifications in this guide apply regardless of narration method and remain the benchmark for all professional distribution platforms.

Amazon does operate a separate AI narration programme called KDP Virtual Voice, which uses Amazon's own AI voices to narrate books enrolled in KDP. This is invite-only as of early 2026 and does not accept author-supplied audio files. It is a different system from ACX.

If you are producing an audiobook with AI narration and want Audible distribution, monitor ACX's policy page directly for any updates. In the meantime, platforms that openly accept AI-narrated audio include Spotify for Authors, Apple Books (via the Digital Narration programme), Google Play Books, Kobo Writing Life, and INaudio (which distributes to 30+ platforms). See the complete AI audiobook distribution guide for step-by-step instructions on each platform.