ACX Rejected Your Audiobook? Fix Every Issue

If ACX rejected your audiobook for RMS, peak, or noise floor, here is the fix. ACX applies two layers of review: an automated spec analysis that measures your file against 8 technical requirements, and a human quality review where a reviewer listens to the audio.[1] Automated spec failures have deterministic fixes. You change a number, the file passes. Human quality rejections require you to re-record or re-edit. This guide covers every rejection type, what causes it, and how to resolve it.

From rejection email to fix, without uploading your file anywhere. ChapterPass processes files locally in your browser. If you are working under a publisher non-disclosure agreement or on an unreleased manuscript, your audio does not leave your device at any point in the fix.

ACX Audio Submission Requirements

ACX measures every uploaded file against three critical specs: RMS loudness between -23 and -18 dBFS, true peak below -3 dBFS, and noise floor below -60 dBFS.[1] Format, sample rate, channel count, and silence padding are also checked. If any measurement falls outside its allowed range, the file is rejected and you receive an email listing which specs failed. For the full list with explanations and cross-platform comparison, see ACX audio requirements (RMS, peak, noise floor).

Files that pass automated analysis move to human quality review. A reviewer listens and evaluates for audible problems: clicks, echo, pacing, manuscript accuracy, and overall production quality. If the reviewer finds issues, you receive a rejection with notes describing what they heard.

How to tell which rejection you got: automated rejections list specific measurements (e.g., "RMS level outside -23 to -18 dBFS"). Quality rejections describe what the reviewer heard (e.g., "excessive background noise" or "audible clicks"). Some rejection emails contain both. If yours does, fix the automated spec issues first, then address the quality notes. There is no point re-recording for quality if the file will fail on specs again.

ChapterPass handles all automated spec adjustments. It cannot fix human quality review failures.

ACX Rejected: Loudness

Loudness rejection is the most common automated failure. Loudness is the average level of your narration across the file, measured as RMS. Most home recordings land between -30 and -40 dBFS RMS. ACX requires -23 to -18 dBFS.[1]

Why -20 dBFS is the target. The allowed range is 5 dB wide. Targeting -20 dBFS gives you 2 dB of margin toward the loud limit and 3 dB toward the quiet limit, which absorbs natural variation between chapters without risking a failure at the edges.

The manual fix. Three steps, in this order:

1. Compress to reduce the gap between your loudest and quietest passages. A ratio of 2:1 to 4:1 with a threshold around -20 dB is a common starting point.[4] This brings quiet sections up without clipping loud ones.
2. Normalise to bring the overall level to your target RMS. Aim for -20 dBFS.
3. Limit peaks. After raising the overall level, some peaks will exceed -3 dBFS. Apply a limiter with a ceiling below -3 dBFS.

The catch. Raising loudness raises everything, including your noise floor. If your raw recording has a noise floor of -64 dBFS and you add 8 dB of gain, the noise floor becomes -56 dBFS, which fails. This is why loudness and noise floor rejections often arrive together, and why you must fix noise floor before adjusting loudness.

A common mistake. Narrators often normalise peaks to -3 dBFS and assume that fixes loudness. Peak normalisation sets the loudest single sample to the target, but RMS measures the average level across the entire file. A file can have a peak at -3 dBFS and an RMS of -35 dBFS if the narration is mostly quiet with occasional loud moments. Compression is what brings the average up. Normalisation alone rarely solves a loudness rejection.

ACX Rejected: Noise Floor

Noise floor is the number one pain point for narrators mastering at home. The problem is not the noise itself. The problem is that boosting loudness to reach -23 to -18 dBFS also boosts the noise above -60 dBFS.

The gain budget. Think of it as a budget. If you need 10 dB of gain to reach your loudness target, your raw noise floor must be below -70 dBFS. If it is -65 dBFS, you are already 5 dB short. Before you start mastering, measure your raw noise floor and calculate how much gain you will need. If the numbers do not add up, no amount of post-processing will fix it cleanly.

Noise reduction vs gating. Broadband noise reduction (like Audacity's Noise Reduction effect[3]) analyses a noise profile and subtracts it from the entire file, including sections where you are speaking. At moderate settings, it works. At aggressive settings, it creates metallic, underwater artefacts that will fail human quality review.

Speech-aware gating takes a different approach. It reduces the volume only during silent sections, where noise is audible, and leaves your voice untouched. The noise is still there under your speech, but listeners cannot hear it because your voice masks it.

Gating cannot remove noise that is present under speech. If your recording environment is noisy enough that you can hear background sound while speaking, improving the space will give you better results than any post-processing.

For the full walkthrough of why noise floor fails after mastering and how gating solves it, see audiobook noise floor guide.

ACX Rejected: Peak

True peak is the most misunderstood spec. Most audio software shows you sample peaks: the loudest individual sample in your file. But when that digital audio is converted back to an analogue signal for playback, the actual waveform can peak higher than any single sample. True peak measurement uses oversampling to detect these inter-sample peaks that a standard meter misses.[2]

The gap. The difference between sample peak and true peak is typically 0.5 to 1 dB, but can reach up to 3 dB in extreme cases. A file that reads -3.2 dB on a sample peak meter could actually exceed -3 dBFS true peak and fail ACX.

The fix. Use a true-peak-aware limiter with a ceiling set below -3 dBFS. A standard limiter catches sample peaks only. A true-peak-aware limiter oversamples internally and catches inter-sample peaks before they exceed the ceiling.

Format, Sample Rate, and Channel Rejections

These are the simplest rejections to fix, but they still account for a significant share of failures.

Format. ACX requires MP3 at 192 kbps CBR (constant bit rate).[1] Files encoded as VBR (variable bit rate), or submitted as WAV, FLAC, or M4A, will be rejected. Export from your DAW as MP3, set to CBR, 192 kbps or higher.

Sample rate. ACX requires 44,100 Hz. Files recorded at 48,000 Hz (common in video production) must be converted. Resample in your DAW before exporting. Do not let the MP3 encoder handle the conversion, as some encoders introduce artefacts during resampling.

Channels. ACX requires mono. If your DAW records in stereo by default, mix down to mono before exporting. A stereo file of spoken word is two identical channels, doubling the file size for no benefit.

Silence padding. ACX requires 0.5 to 1 second of room tone at the head and 1 to 5 seconds at the tail.[1] The silence must be room tone, not digital silence (absolute zero). Digital silence creates an audible click at the transition. If your file was rejected for silence, measure the existing silence duration in your DAW and adjust. Be consistent across all chapters: pick the same head and tail lengths and apply them everywhere.

For the complete spec table and how these values compare across platforms, see ACX audio requirements (RMS, peak, noise floor).

What Order to Fix Multiple Issues

If your file failed on more than one spec, the order you fix them in can make the process easier. Each processing step changes the measurements that follow, so a sensible sequence avoids rework. This is not a rigid rule, but a recommended workflow:

1. Sample rate and channels. Convert to 44,100 Hz mono first, while the file is still WAV. This is not MP3 encoding yet (that happens at step 6). All subsequent measurements depend on the file being at the final sample rate and channel count.
2. Noise floor. Apply gating or noise management before touching loudness. If you boost loudness first, you lock in a higher noise floor.
3. Loudness (RMS). Compress and normalise to -20 dBFS. Now that noise is managed, the gain increase will not push noise above -60 dBFS.
4. True peak. Apply a true-peak-aware limiter with ceiling below -3 dBFS. Loudness processing may have introduced new peaks, so this step comes after.
5. Silence padding. Add 0.5 to 1 second of room tone at the head and 1 to 5 seconds at the tail. Do this last, after all level processing is complete.
6. Encode. If you have been working in WAV, export the final MP3 as the last step. Encoding is lossy, so you only want to do it once.

Why order matters, in practice. If you compress and normalise a file to -20 dBFS RMS before managing the noise floor, the noise is now baked into the louder signal. Gating after amplification will still help in silent sections, but the noise under speech is permanently louder. Handling noise first, while the signal is still quiet, means you are managing a smaller problem. The same principle applies to peaks: limiting before loudness adjustment wastes headroom, because the loudness step may push levels up again and create new peaks that need re-limiting.

Quality Rejections

Human quality review catches problems that automated spec measurement does not. These require you to re-edit or re-record the affected sections.

• Clicks, pops, and mouth noise. Audible artefacts from lip and tongue movement. Remove with a de-clicker plugin or manual editing in your DAW. Audacity audiobook mastering has a click removal tool.
• Room echo and reverberation. Sound bouncing off walls, floors, and ceilings. The only real fix is recording in a treated space. Post-processing de-reverb tools exist but often introduce artefacts.
• Pacing and breath timing. Unnatural pauses, rushed sections, or audible gasping between sentences. This is a performance issue, not a technical one.
• Opening and closing credits. ACX requires specific credit content: title, author name, narrator name, and publisher. Missing or incorrect credits will fail review.
• Manuscript accuracy. The narration must match the approved manuscript. Skipped paragraphs, changed words, or improvised sections will be flagged.
• Inconsistent tone. Chapters recorded on different days with different mic positions, gain settings, or room conditions can sound noticeably different. Reviewers listen for consistency across the full audiobook.

Start with a clean, well-edited recording, and the technical mastering will take care of itself.

How to Resubmit

1. Read the rejection email carefully. Note whether it cites specific measurements (automated) or describes audible problems (quality review).
2. Fix quality issues first if both types are present. There is no point perfecting the specs on audio that will fail quality review.
3. Fix automated spec issues in order. Follow the processing order in the previous section: format, noise floor, loudness, peaks, silence, encode.
4. Verify your fixed files. Load them into a meter that shows RMS, true peak, and noise floor. Confirm all values are within range before uploading.
5. Upload the corrected files to ACX, replacing the rejected ones. ACX allows you to replace individual chapter files without resubmitting the entire audiobook.
6. Resubmit for review. The review process starts over from the beginning. Make sure every chapter passes, not just the ones that were flagged.

Common Questions

Your first file is free at ChapterPass.