Audiobook Loudness: The Complete Guide to RMS, LUFS, and dBFS

Loudness is the most fundamental technical requirement for audiobook distribution. ACX requires RMS loudness between -23 and -18 dBFS on every chapter file. Get it wrong and your audiobook gets rejected.[1] But "getting it right" requires understanding what loudness actually means in digital audio, how different measurement standards compare, why your tools might disagree with ACX's check, and how the interaction between loudness and noise floor can create cascading failures.

This guide covers everything an audiobook producer needs to know: the measurement standards (RMS, LUFS, peak), the specific ACX requirements, how ACX's automated check works internally, and the techniques to hit the target consistently across every chapter.

What Is Loudness in Digital Audio?

Loudness is not volume. Volume is the physical level at which audio plays back, the number on your amplifier's dial. Loudness is a property of the audio signal itself, measured against a digital reference point called dBFS (decibels relative to full scale). 0 dBFS is the maximum level a digital system can represent. Every practical measurement is a negative number: -20 dBFS is quieter than -10 dBFS.[2]

What Are the Three Ways to Measure Loudness?

Peak Level

Peak level is the single loudest sample in a file, the highest point the waveform reaches. It tells you the maximum instantaneous level but nothing about how loud the file sounds overall.

True peak goes further by calculating the actual analog waveform between digital samples using 4× oversampling per the ITU-R BS.1770 standard.[2] Inter-sample peaks can exceed the level of any individual sample by 0.5–1.5 dB, which is why ACX specifies a true peak limit rather than a sample peak limit. ACX requires true peak below -3 dBFS.

RMS (Root Mean Square)

RMS is a measurement of average signal power across an audio file's full duration.[1] The calculation squares every sample value, averages the squares, then takes the square root, giving disproportionate weight to louder samples. RMS includes both speech and silence, so a file with long pauses measures lower than a densely narrated file even if the speech portions are identical.

ACX requires RMS between -23 and -18 dBFS. ACX measures full-file RMS of the decoded MP3, including head and tail silence. This means silence padding pulls your measurement down, which is significant for short chapters with generous tail padding.

LUFS (Loudness Units Full Scale)

LUFS applies a frequency weighting curve (K-weighting) that boosts 2–4 kHz where human hearing is most sensitive and attenuates frequencies below 200 Hz, making it more perceptually accurate than raw RMS.[3] LUFS also uses gating that excludes momentary loudness below -70 LUFS, so silence periods don't pull the measurement down as much.

LUFS is the broadcast standard (EBU R128, ITU-R BS.1770), used by Spotify and YouTube. ACX uses RMS, not LUFS, an important distinction. For spoken word, RMS and LUFS typically track within 1–2 dB, but they're not interchangeable. If you use a LUFS meter for ACX content, your file may land outside the required range.

What Are the Exact ACX Loudness Requirements?

The -23 to -18 dBFS RMS range is a five-decibel window.[1] This range ensures listeners don't need to adjust volume between chapters or between audiobooks. It's deliberately lower than music (commonly -14 to -8 LUFS) and podcasts (-16 to -14 LUFS). Audiobooks are consumed for hours at a time, often in quiet environments, so a lower loudness target reduces listening fatigue.

The lower bound of -23 overlaps with the EBU R128 broadcast standard target of -23 LUFS, but the two standards use different measurement systems: ACX uses RMS in dBFS while EBU R128 uses LUFS with K-weighting.[3]

How Does ACX's Automated Check Measure Your Files?

Understanding how ACX measures helps you match their result.

Full-file measurement: ACX measures the RMS of the entire MP3 file: head silence, speech, pauses, tail silence. The silence portions pull RMS down. A 10-minute chapter with 5 seconds of tail silence sees a negligible effect, but a 3-minute chapter with 5 seconds of tail silence can see a 1 dB drop. This is why targeting -20 dBFS gives essential margin.

MP3 decode then measure: ACX decodes the submitted MP3 to PCM and measures the resulting audio. The decoded signal differs slightly from your source WAV due to lossy compression, so measurements can vary by 0.3–0.5 dB between your source and the delivered MP3.

Per-file, not per-audiobook: Each chapter is measured independently. A file at -24 dBFS fails even if all other files are at -19 dBFS. There's no averaging across files.

Why Does Your Measurement Differ from ACX's?

Tool differences: Not all RMS meters calculate identically. Some use sliding windows; others calculate a single full-file value. Window size, preprocessing, and floating-point precision can produce variations of 0.3–0.5 dB. The solution is margin: target -20 dBFS so ±0.5 dB of variance still keeps you in range.

Measuring the wrong file: The most common cause of discrepancy. If you measured your project or WAV but submitted an MP3, the values may differ. Always measure the actual MP3 you're submitting.

How Do You Measure Audiobook Loudness?

Audacity (Free)

Audacity with the ACX Check plugin measures RMS, peak, and noise floor in one pass. Note: Audacity v3.x supports both LUFS and RMS normalisation modes. Use the RMS mode for ACX, since that's what ACX measures.[1] The LUFS mode applies frequency weighting that produces a different number.

Select all audio → Analyze → ACX Check (or Analyze → Contrast for manual RMS reading).

Youlean Loudness Meter (Free)

A free VST/AU plugin and standalone application that measures both RMS and LUFS in real time, plus true peak using the BS.1770 standard.[2] Works in any DAW that supports plugins.

Professional Tools

Adobe Audition's Amplitude Statistics provides RMS, peak, and true peak. iZotope RX's Loudness module gives the most complete metering in a single view, with batch measurement for processing multiple chapters.

How Do You Adjust Audiobook Loudness?

The goal is to bring your full-file RMS to approximately -20 dBFS, the centre of ACX's range, with 2 dB of margin on each side. The how to fix audiobook loudness guide covers the specific adjustment steps in a more hands-on format.

What Is the Correct Signal Chain?

Loudness adjustment is part of a signal chain where order matters:

1. High-pass filter (80 Hz): removes low-frequency energy that inflates RMS
2. Noise treatment: address background noise before raising levels, since gain amplifies noise linearly
3. Compression: narrow dynamic range so gain adjustment affects the whole signal evenly
4. Gain adjustment: raise or lower the overall level to hit the RMS target
5. True-peak limiting: catch any peaks that exceed -3 dBFS after gain adjustment
6. Verification: measure the final output file

See the full mastering guide for detailed settings at each step.

What Compression Settings Work for Spoken Word?

Compression doesn't make audio louder. It makes quiet parts closer to loud parts, so less gain is needed to bring the signal into range.

Recommended settings:

• Ratio: 2:1 to 3:1
• Threshold: -22 to -18 dBFS
• Attack: 15–25 ms
• Release: 200–500 ms
• Knee: Soft (gradual onset sounds more natural)

A ratio above 4:1 squashes the natural dynamic range of speech, making narration sound flat and fatiguing. If you need heavy compression to hit the target, the recording level is the problem, not the mastering.

Should You Use RMS Normalisation or LUFS Normalisation?

For ACX, use RMS normalisation if your tool supports it (Audacity v3.x does). LUFS normalisation adjusts to a target LUFS level, which introduces a 1–2 dB offset for spoken word. A file normalised to -20 LUFS will typically measure around -22 to -23 dBFS RMS, barely within ACX's range. If you must use a LUFS meter, target -22 LUFS for ACX content.

Peak normalisation is unreliable for hitting an RMS target because the relationship between peak and RMS depends on the dynamic range of the content.

How Do You Achieve Chapter-to-Chapter Consistency?

ACX doesn't just check individual chapter loudness. It also evaluates consistency across chapters. Large RMS variations suggest inconsistent recording or processing.

What causes inconsistency: Recording sessions on different days with different mic positions or room conditions. Per-chapter manual processing with slightly different settings. Mixed source material from different microphones or locations.

How to achieve consistency: Record with the same setup every session. Batch process all chapters with identical mastering settings. Check that every chapter's RMS is within 1 dB of your target. If one chapter is significantly different, investigate the recording conditions before adjusting.

What Loudness Mistakes Should You Avoid?

Measuring the wrong file. Always measure the final MP3, not the project file or intermediate WAV. MP3 encoding can shift RMS by up to 0.5 dB and true peak by 0.5–1 dB. If your file is getting rejected specifically for loudness, the RMS loudness rejection guide walks through the most common causes and fixes.

Normalising to peak instead of RMS. Peak normalisation sets the highest peak to a target level. Two files peak-normalised to -3 dBFS can have completely different RMS values.

Using music loudness targets. Music targets -14 to -8 LUFS. Podcast targets -16 to -14 LUFS. Audiobook targets -23 to -18 dBFS RMS. These are different formats with different listening contexts.

Ignoring the interaction with noise floor. Every decibel of gain also raises your noise floor by one decibel. A raw recording at -62 dBFS noise floor with 8 dB of gain becomes -54 dBFS, above ACX's -60 dBFS threshold.

Over-compressing. If you can't hit -20 dBFS without a compression ratio above 4:1, the recording level is the problem. Re-record with higher input gain rather than crushing dynamics in post-production.

What Tools Measure Audiobook Loudness?

For a detailed comparison, see best audiobook mastering tools compared. For common questions about loudness and other ACX requirements, see the FAQ or the ACX platform page.