How to Fix Audiobook Loudness: Too Quiet, Too Loud, and Inconsistent Chapters

Your audiobook's loudness needs to land between -23 and -18 dBFS RMS to pass ACX's technical review.[1] That sounds straightforward until you measure your files and find Chapter 3 at -28 dBFS, Chapter 7 at -17.5 dBFS, and Chapter 12 at -24 dBFS. Some are too quiet, one is borderline too loud, and nothing is consistent.

This guide covers three specific problems and their fixes: audiobooks that are too quiet, audiobooks that are too loud, and audiobooks with inconsistent loudness between chapters. Each section includes exact steps, settings, and verification.

How Do You Fix an Audiobook That's Too Quiet?

This is the most common loudness problem. Most home recordings land between -30 and -40 dBFS RMS because narrators set conservative input levels to avoid clipping.

Why Is Your Recording Too Quiet?

• Input gain too low. The most common cause. Your audio interface gain is set so low that the signal never reaches adequate levels.
• Microphone too far. Every doubling of mic distance reduces level by approximately 6 dB. Recording at 16 inches instead of 8 inches costs 6 dB.
• Low-sensitivity mic without enough gain. Dynamic microphones like the SM7B require more gain than condensers. Without a preamp (Cloudlifter, FetHead), they often produce insufficient levels.
• Naturally soft speaking style. Some narrators speak softly, producing lower signal levels.

Step-by-Step Fix

Step 1: Measure current RMS. Open your final MP3 in your DAW. Select all audio and measure full-file RMS. The difference between this and -20 dBFS tells you how much processing you need. ACX measures RMS (root mean square of sample values), which represents average signal power, not LUFS, which applies K-weighting.[2]

Step 2: Address noise floor first. Every dB of gain you add also raises the noise floor by 1 dB.[2] If your noise floor is -62 dBFS and you need 8 dB of gain, your noise floor will become -54 dBFS, above the -60 dBFS limit.[1]

Apply noise treatment before gain adjustment:

• High-pass filter at 80 Hz (removes low-frequency rumble)
• Noise gate during pauses (replaces noisy silence with clean room tone)
• Light spectral noise reduction if needed

Step 3: Apply compression. Compression narrows the dynamic range, reducing how much gain you need. For spoken word:

• Ratio: 2:1 to 3:1 (above 4:1 destroys speech dynamics)[3]
• Threshold: approximately -24 dBFS
• Attack: 20 ms
• Release: 300 ms
• Knee: Soft

Don't apply makeup gain on the compressor. Handle gain separately.

Step 4: Adjust gain to hit -20 dBFS. With compression applied, your RMS should be higher but still below target. Apply positive gain to reach approximately -20 dBFS RMS. Re-measure after adjustment.

Step 5: Apply true-peak limiting. After gain increase, some peaks will exceed the ceiling. Apply a brickwall limiter with a true-peak ceiling of -3.1 dBFS. Use a limiter with true-peak detection, not just sample-peak. Inter-sample peaks can exceed sample peaks by 0.5–1 dB.[3]

Step 6: Export and verify the MP3. Export as MP3, 192 kbps CBR, 44100 Hz, mono. Measure the final MP3:

• RMS: between -23 and -18 dBFS ✓
• True peak: below -3 dBFS ✓
• Noise floor: below -60 dBFS ✓

If noise floor fails, go back to Step 2 with more aggressive noise treatment. The noise floor guide covers room-specific solutions.

How Do You Fix an Audiobook That's Too Loud?

Less common but equally problematic. Files with RMS above -18 dBFS get rejected just like files below -23 dBFS.

Why Is Your Recording Too Loud?

• Music mastering settings. Music targets -14 to -8 LUFS. Applying these settings to audiobooks produces files 4–10 dB too loud.
• Over-compression. Ratios above 4:1 flatten dynamics and raise average level above range.
• Aggressive normalisation. Normalising to -16 or -14 dBFS (common podcast targets) pushes past the ceiling.
• Stacked processing. Multiple compression stages, each adding gain.

Step-by-Step Fix

Step 1: Measure current RMS. Determine how far above -20 dBFS you are.

Step 2: Reduce compression. If you used a ratio above 3:1, reduce to 2:1. Disable makeup gain and let subsequent gain adjustment handle targeting.

Step 3: Apply negative gain. Attenuate overall level to bring RMS down to approximately -20 dBFS.

Step 4: Verify all specs. Reducing level also reduces peak levels and effective noise floor. Re-measure everything.

How Do You Fix Inconsistent Loudness Between Chapters?

Individual chapters pass, but the spread between quietest and loudest is 5 dB or more. ACX's human review may flag this as inconsistent audio quality, and listeners will notice volume jumps between chapters.

Why Do Chapters Have Different Loudness?

• Recording sessions on different days. Room conditions, mic placement, and vocal energy change.
• Vocal performance variation. Dialogue chapters have different dynamics than exposition.
• Manual processing with slightly different settings. Hand-mastering introduces cumulative drift.
• Mixed source material. Different rooms or microphones.

Step-by-Step Fix

Step 1: Measure all chapters. Create a spreadsheet listing every chapter file with its full-file RMS to reveal the pattern.

Step 2: Identify outlier causes. If Chapter 3 is 2.5 dB quieter than surrounding chapters, investigate the source: different session, more whispered dialogue, different noise floor.

Step 3: Batch re-process from source files. The most reliable fix: batch processing all chapters from source through a single, identical mastering chain. Apply the same high-pass filter, compressor settings, gain target, limiter ceiling, and encoding settings to every file.

Most DAWs support batch processing: Audacity's Macro feature, Adobe Audition's Batch Processing, Reaper's batch rendering. The Audacity tutorial and the mastering tools comparison cover batch options in detail.

Step 4: Verify the full set. After batch processing, re-measure all chapters. The spread should be 1–2 dB at most.

How Do You Achieve Consistent Loudness Across All Chapters?

Individual chapters that each pass ACX's automated check can still produce a poor listener experience if their loudness varies significantly from one to the next. When a listener moves from Chapter 4 to Chapter 5 and has to reach for the volume control, that's an ACX human review flag and a listener complaint waiting to happen.

What "consistent" means in practice

ACX requires each file to measure between -23 and -18 dBFS RMS. A chapter at -18.5 dBFS and another at -22.5 dBFS both pass the automated check individually. But the spread between them is 4 dB. Listeners will notice. The better target is all chapters within 1 dB of each other, ideally within 0.5 dB, all clustered around -20 dBFS.

Why chapters drift apart

Different recording sessions. Vocal energy, mic placement, and room conditions vary between sessions. A chapter recorded on a quiet Sunday morning will measure differently than one recorded on a noisy weekday afternoon, even with identical settings.

Performance variation. Action-heavy chapters with dialogue, raised voices, and dramatic pacing produce different average levels than quiet introspective chapters heavy on internal monologue. This is expected and natural, but it requires the mastering chain to compensate.

Manual processing drift. When mastering chapters individually by hand, small differences in how you set thresholds and gain across 30 or 40 files accumulate into audible inconsistency.

Mixed source material. If chapters were recorded in different rooms or on different microphones, their tonal character and noise profiles differ, and their starting levels before mastering differ too.

The batch processing approach

The most reliable fix for chapter consistency is also the simplest: run all chapters through an identical automated mastering chain rather than processing them individually by hand.

Most DAWs support batch or macro processing. In Audacity, the Macro feature (Edit > Macros) lets you define a sequence of processing steps and apply them to a folder of files. In REAPER, you can set up a project with a consistent processing chain and batch render multiple files. Adobe Audition has a dedicated Batch Processing panel.

The principle is the same regardless of tool: define your signal chain once (high-pass filter, compression settings, loudness target, limiter ceiling, export format) and apply it identically to every file. The result is chapters that are consistent by design rather than by manual approximation.

ChapterPass applies this approach automatically. Upload all chapters and each file passes through the same mastering chain, producing consistent output across the full set.

Measuring consistency before submission

After processing all chapters, don't just spot-check two or three files. Measure every chapter's full-file RMS and list them. If any file is more than 1 dB from the others, investigate the source before submitting.

The FFmpeg loudnorm command outputs RMS per file. Run it as a batch across all your MP3s:

ffmpeg -i chapter_01.mp3 -af loudnorm=print_format=json -f null -

Run this for each file and compare the input_i (integrated loudness) values. Files with significantly different values from the set median need investigation, either a mastering chain adjustment or a source recording problem that mastering can't fully compensate for.

When chapter content causes legitimate variation

Some variation between chapters is acceptable. A chapter of whispered dialogue will measure differently than a chapter of tense action even with identical mastering settings, because the underlying dynamics differ. The mastering chain narrows the gap but doesn't eliminate it entirely.

ACX's automated check passes both as long as each falls within -23 to -18 dBFS. The human review threshold is higher: reviewers listen for jarring jumps, not small variations. A spread of 1 to 2 dB across all chapters is unlikely to trigger a human review flag. A 4 to 6 dB spread will.

What Are the Most Common Loudness Mistakes?

Most loudness problems come from one of five mistakes. Each has a different cause and a different fix.

Mistake 1: Normalising to a peak target instead of an RMS target

Normalisation to peak (setting the highest sample to -3 dBFS, for example) is common in music workflows and in basic DAW presets. For audiobooks, it produces inconsistent results. A quiet chapter with a single loud transient gets normalised based on that transient, leaving the overall level far too low. The peak hits -3 dBFS but the RMS might be -30 dBFS.

Fix: Always target RMS, not peak. Normalise to -20 dBFS RMS, then apply a peak limiter to catch anything above -3 dBFS. Never use peak normalisation as a loudness step for audiobooks.[1]

Mistake 2: Over-compression flattening all dynamics

Compression ratios above 4:1 destroy the natural dynamics of speech.[3] The result sounds pumped, unnatural, and fatiguing. It also pushes the RMS higher than intended because transients no longer pull the average down.

Narrators sometimes apply heavy compression chasing a loud, punchy sound or trying to eliminate all variation. The result is audio that's technically too loud and artistically unpleasant.

Fix: Use 2:1 to 3:1 ratio for spoken word. Apply makeup gain separately from compression. If your RMS is still below target after gentle compression, add gain rather than increasing compression ratio.

Mistake 3: Skipping the noise floor check after boosting gain

Every dB of gain you add raises the noise floor by 1 dB.[2] A recording with a noise floor of -63 dBFS that needs 8 dB of gain to reach -20 dBFS RMS will have a noise floor of -55 dBFS after processing. That's well above the -60 dBFS limit.

Narrators often check noise floor on the raw recording and assume it's fine. They don't re-check after the loudness adjustment that was the whole point of their processing.

Fix: Always measure noise floor on the final MP3, not on the raw recording. If it fails, go back and apply more noise treatment before the gain step, or reduce the amount of gain you're applying.

Mistake 4: Targeting -18 dBFS instead of -20 dBFS

The ACX range is -23 to -18 dBFS. Some narrators target -18 dBFS because it's the loudest allowed and they believe louder is better. The problem is that any measurement uncertainty or processing step after the loudness adjustment can push the file above -18 dBFS, causing a failure.

Targeting the centre of the range (-20 dBFS) gives 2 dB of margin on each side. A file at -20 dBFS can drift 2 dB in either direction and still pass. A file at -18.1 dBFS fails.

Fix: Target -20 dBFS RMS. The 2 dB margin is there for a reason.

Mistake 5: Applying the same settings from a previous project without re-checking

A mastering chain calibrated for one narrator's recording environment and vocal character won't produce identical RMS results on a different narrator's files. The compressor threshold, attack, and release settings that work perfectly for a soft-voiced narrator recording in a carpeted bedroom will produce different results for a projected theatrical narrator in a livelier room.

Narrators and producers who work across multiple projects often apply saved presets without verifying the output of each new project.

Fix: Always measure the output of every project before submission. Presets are a starting point, not a guarantee. Run your final measurement on the actual MP3 files you plan to submit.

Why Does the Loudness Adjustment Order Matter?

The processing sequence affects results. Here's the correct order:[1]

1. High-pass filter (80 Hz): Low-frequency energy inflates RMS without contributing to intelligibility.
2. Noise treatment: Reduce noise before raising levels. Gain amplifies noise 1:1.[2]
3. Compression: Narrow dynamic range before applying gain. Less gain needed means less noise amplification.
4. Gain adjustment: Raise (or lower) overall level to hit -20 dBFS RMS.
5. True-peak limiting: Catch any peaks exceeding -3 dBFS after gain adjustment.
6. Silence padding: Add head and tail silence.
7. MP3 encoding: Convert to 192 kbps CBR.
8. Verification: Measure the final MP3 against all ACX specifications.

The complete mastering guide covers detailed settings at every stage.

When Should You Re-Record Instead of Fix?

Sometimes the problem is too severe for post-production:

• Raw recording below -45 dBFS RMS. You'd need 25+ dB of gain, which amplifies noise and introduces artefacts even with aggressive treatment.
• Wildly inconsistent recording levels. Chapters varying by more than 10 dB in raw RMS indicate too-different recording conditions.
• Compression artefacts from previous mastering. Over-compressed audio can't be uncompressed. You need the pre-mastered source.

Re-recording with proper gain staging is faster than rescuing unsuitable material. Set input gain so normal narration peaks around -12 to -6 dBFS, record in a quiet environment, and keep consistent mic distance across all sessions.

Quick Reference: Loudness Targets