What ACX Actually Checks When You Submit an Audiobook

Most guides about ACX submission focus on the technical specs: loudness, peaks, noise floor. Those matter, and they're the reason a lot of audiobooks get rejected on the first try. But they're only half the story. ACX runs two independent review layers on every audiobook: an automated technical check and a human quality review. Both must pass.[1]

What Are the Two Layers of ACX Review?

Every audiobook submitted to Audible through ACX goes through two separate checks before it can go on sale:

1. Automated technical checks. A system measures your audio files against 8 measurable specifications: RMS loudness (-23 to -18 dBFS), true peak (below -3 dBFS), noise floor (below -60 dBFS), sample rate (44100 Hz), mono channel, MP3 CBR 192 kbps, and head/tail silence.[1]
2. Human quality review. An ACX QA reviewer listens to your audio and evaluates recording quality, narration consistency, and manuscript accuracy.

A technically perfect file with audible mouth clicks will get rejected. A beautifully recorded chapter with the wrong loudness will also get rejected. You need to clear both hurdles.

What Technical Specs Does ACX Check Automatically?

ACX's automated system measures every chapter file against 8 measurable specifications. Files that fail any single spec get rejected. The ACX audio requirements guide covers each spec in full technical detail.

Volume levels. Each chapter's RMS loudness must fall between -23 and -18 dBFS. All chapters in a single audiobook need to be consistent. If one chapter is noticeably louder or quieter than the others, it gets flagged.[1]

Peak levels. No true peak in your audio can exceed -3 dBFS. This includes inter-sample peaks that occur between digital samples, which standard audio meters often miss. The audio mastering glossary defines true peak and explains why it differs from sample peak. Peaks above -3 dBFS cause distortion on playback devices, especially earbuds and phone speakers.[2]

Noise floor. When nobody is speaking, the background noise must stay below -60 dBFS. This catches room noise, HVAC hum, computer fans, and electrical interference. ACX measures this during the silent gaps between sentences and paragraphs. Noise floor violations are among the most common ACX rejection reasons.[1]

File format. Files must be mono MP3, 44100 Hz sample rate, 192 kbps constant bit rate (CBR). Other formats, variable bit rate encoding, stereo files, and non-standard sample rates all result in immediate rejection.

Silence padding. Each chapter needs 0.5 to 1 second of silence at the beginning and 1 to 5 seconds at the end. Too short, too long, or digital silence (absolute zero instead of room tone) all fail.

Consistency. All chapters must have consistent loudness and noise characteristics. Batch processing through the same mastering chain is the standard approach for ensuring this.

This is the layer ChapterPass handles. It adjusts your audio to meet every one of these measurable specs and verifies each file before you download it.

What Causes Automated Check Failures?

Understanding the common failure modes helps you avoid them. These are the technical rejections narrators hit most often, ordered roughly by frequency:

RMS too low (most common). Raw home studio recordings typically land between -28 and -35 dBFS RMS, well below the -23 dBFS minimum. The fix is normalisation plus compression, but doing this wrong creates a cascade: boosting a quiet recording amplifies the noise floor along with the speech.[3]

Noise floor too high. The second most common failure, and often a consequence of fixing the first. Room noise, HVAC, electrical hum, and computer fans all contribute. The threshold is -60 dBFS, which is quieter than most people realise. In a typical home office with a computer running, the ambient noise might measure -50 to -55 dBFS. That fails.

True peak exceeding -3 dBFS. This happens when narrators normalise loudness without applying a limiter afterward. A chapter can have an RMS of -20 dBFS (within range) while a single plosive or emphasis peak hits -1 dBFS. A true peak limiter set to -3 dBFS is the standard fix, but it must be a true-peak limiter, not a regular peak limiter.[2]

Wrong format. Submitting WAV instead of MP3, variable bit rate instead of constant, stereo instead of mono, or the wrong sample rate. These are binary pass/fail and easy to catch before submission.

What Does ACX's Human Quality Review Check?

After your files pass the automated checks, an ACX QA reviewer listens to your audio. This review catches recording and performance issues that automated systems cannot detect.

Clicks, pops, and mouth noise. Recording artefacts that automated checks don't catch. A tongue click, a lip smack, a pop from a plosive. These need to be cleaned up in your editing software.

Room echo and reverb. If your recording space isn't treated acoustically, your narration will have a noticeable room sound. Small rooms with hard walls are the worst offenders. ACX reviewers listen for this because it degrades the listening experience, especially on headphones.

Pacing and consistency. Your narration should sound like it was recorded in one session, even if it wasn't. Large shifts in energy, pace, or tone between chapters get flagged.

Opening and closing credits. ACX requires your audiobook to start with opening credits (title, author, narrator) and end with closing credits. Missing credits is a straightforward rejection.

Manuscript accuracy. The reviewer checks that your narration matches the written book. Skipped paragraphs, wrong words, and improvised sections can all trigger a rejection.

Human narration. ACX requires human narration. The only exception is ACX's Voice Replica programme (beta since September 2024), which allows narrators to create and monetise AI replicas of their own voice with explicit consent. Outside of that programme, your audiobook must be narrated by a human.

No tool can handle these for you. They depend entirely on your recording quality, your editing, and your performance.

What Does a Typical Rejection Look Like?

When ACX rejects your audiobook, you receive a notification identifying which files failed and why. The feedback varies in specificity:

Automated rejections are precise. You'll get a message identifying the exact spec that failed: "File Chapter03.mp3: RMS level -25.2 dBFS is below the minimum -23 dBFS" or "True peak exceeds -3 dBFS." These are straightforward to fix because the problem is quantified.

Human review rejections are less specific. Common feedback includes:

• "Excessive mouth clicks": clean these in your editing software, then re-master
• "Room tone inconsistency": chapters recorded on different days in different conditions
• "Audio artefacts present": often caused by aggressive noise reduction, clipping, or lossy-to-lossy format conversion
• "Missing opening/closing credits": record and add them as the first and last files

The rejection cycle is the most time-consuming part of ACX submission for many narrators. Each round trip can take several days to a week. Getting the technical specs right on the first pass eliminates half the potential rejection reasons.

What Does the Human Review Actually Listen For in Practice?

The ACX QA team's scope is narrower than most narrators assume. ACX is explicit that its QA team listens for submission requirements but does not conduct an end-to-end review of the audiobook. This means reviewers are listening against a checklist, not performing a full artistic evaluation.

What that checklist covers:

Extraneous sounds. The whir of a fan, a refrigerator hum, electrical hiss, mouse clicks, chair squeaks, and air conditioning are all flagged. These are distinguishable from room tone because they fluctuate or have a tonal character. Room tone is neutral and stable. HVAC noise has pitch and rhythm.

Outtakes and false starts left in the file. Narrators who don't edit carefully sometimes leave retakes or stumbled lines in the file. The QA listener catches these because they break the flow of the narration.

Over-processed audio. Aggressive noise reduction and extreme compression create artefacts that sound robotic or clipped. Files that have been processed too heavily fail human review even if they pass every automated spec. There is such a thing as over-mastering.

File structure problems. Each chapter file should contain exactly one chapter or section. According to ACX's own QA guidance, improper file grouping is among the most common human-review failures: multiple chapters in a single file, or a chapter split across files when it doesn't need to be. Long chapters over two hours or 170 MB may be split, and very short chapters under five minutes may be grouped, but the standard is one file per chapter.

Gaps of silence inside a chapter. Internal silence gaps of more than a few seconds flag as editing errors. The standard fix is to replace them with room tone so the ambience is consistent throughout. For more on room tone versus digital silence, see the silence padding guide.

The things that fall outside this scope: vocal performance quality, character distinction, creative narration choices, tonal consistency within a chapter. ACX's advice is that these are the author's and narrator's responsibility to verify before submission, because QA may miss them.

What Happens After ACX Rejects Your Audiobook?

Rejection notifications come from the same portal where you submitted. The feedback identifies which files failed and what category of issue caused the failure. The level of detail varies.

Automated rejections include the specific measurement and which spec it violated. Human review rejections use category labels ("audio artefacts present," "extraneous sounds," "outtakes detected") without timestamps or specific chapter positions.

Fix Only the Flagged Files

You do not need to resubmit every chapter. Only the files that failed need to be corrected and re-uploaded. If three chapters fail noise floor and the rest pass, fix those three.

One important caveat: if the failure was caused by a mastering chain error (a wrong limiter setting, an incorrect normalisation target, a bad export template), that same error may exist in chapters that weren't flagged. Spot-check all chapters processed through the same chain before assuming the unflagged ones are clean.

Fixing a Human Review Rejection

Human review rejections require identifying the specific problem by listening through the rejected chapter on headphones. If the feedback says "extraneous sounds," listen for the noise source, not just the chapter as a whole. If it says "audio artefacts present," this usually means over-processing: undo whatever aggressive noise reduction or compression was applied, re-process with gentler settings, and verify the output.

After fixing the issue, re-master the corrected chapter and verify every spec before re-uploading. Fixing one problem can shift measurements. A chapter re-processed without a limiter after noise reduction changes may now have peaks above -3 dBFS. Always run a full verification pass on any re-processed file.

Resubmission Timelines

ACX's human review takes between 5 and 10 business days per round.[1] There is no published cap on the number of resubmission attempts, but every failed round costs you up to two weeks of sales time. First-submission acceptance is worth working toward precisely because the resubmission cycle compounds: one failure delays your book by two weeks, two failures delay it by a month.

How Do You Pass Both Layers?

You need two things: a clean recording and correct technical mastering. Neither one alone is enough.[1]

For the human review: Record in a quiet, treated space. Use a quality microphone with a pop filter. Edit out mistakes, mouth noise, and any audible artefacts. Record your credits. Keep your narration consistent across sessions. Listen to your final audio on headphones before submitting.

For the technical specs: This is what ChapterPass handles. Upload your edited, clean recordings and the engine adjusts loudness, limits peaks, manages noise floor, converts to the right format, and pads silence. Every file is verified before you download it. The ACX requirements page has the full spec table and explains what ChapterPass automates.

The key insight is that these two layers are independent. A great recording with bad mastering fails. Perfect mastering on a noisy recording also fails. You need both sides handled. See the FAQ for more on how ChapterPass fits into this workflow.