Why audiobooks are quieter than everything else
Loudness is measured in LUFS (loudness units relative to full scale): the closer to zero, the louder. Different worlds aim at different levels:
| Where | Level | Who says so |
|---|---|---|
| Music on Spotify | −14 LUFS | Spotify normalises playback to it1 |
| Podcasts on Apple Podcasts | −16 LUFS ±1, true peak −1 dB | Apple’s audio requirements2 |
| TV and radio in Europe | −23 LUFS | EBU R 1283 |
| Audiobooks on Audible (via ACX) | −23 to −18 dB RMS | ACX’s submission requirements4 |
| Two LibriVox chapters we measured | −27 and −28 LUFS | Our measurement, October 2026 |
ACX measures RMS rather than LUFS, but the gist is plain: audiobooks are allowed to be much quieter than music and podcasts, and volunteer and older recordings often are. Switch from a song to a chapter and you reach for the volume; in a car or on a train, the top of your phone’s range may not be enough.
What the booster does
- Measures the book’s integrated loudness (EBU R 128, as ffmpeg’s meter computes it) on twelve 30-second stretches spread through it. On our test chapters that came within 0.1 dB of measuring every second, in about two seconds instead of minutes.
- Tests the gain. The gain to your target is tried on the same stretches through the limiter, and corrected (up to three times), because a limiter takes a little loudness from lively narration.
- Limits the peaks at −2.5 dBFS, so true peaks stay under −1 dB even after encoding, Apple’s ceiling for spoken audio.2
- Re-encodes once, at the source’s own bitrate, with its chapters, cover and tags.
- Measures the result the same way, so the number you see is the file you get.
Tested
| Book | Before | Target | After (whole file) | True peak |
|---|---|---|---|---|
| LibriVox, 20-minute chapter | −27.0 LUFS | −16 | −16.0 LUFS | −1.4 dB |
| LibriVox, 17-minute chapter, evened out | −28.3 LUFS | −16 | −16.1 LUFS | −1.5 dB |
| Uneven test file (passages 6 dB apart), evened out | −36.8 LUFS | −16 | −15.9 LUFS | −1.5 dB |
Measured with ffmpeg’s EBU R 128 meter over the whole processed file, before encoding; the M4B fixture in our browser test came out at −16.2 LUFS with a −1.2 dB true peak after AAC encoding. The tool was also measured against ffmpeg’s own loudnorm filter: in a single pass it stopped 3 dB short of the target and took more than twice as long, so it isn’t used.
Even out, or just louder?
“Just louder” turns the whole book up by the same amount: the narrator’s performance, loud and soft, is kept exactly. “Even out” also narrows the gap between quiet and loud stretches, over a window of about 15 seconds, so a whispered chapter or a narrator who drifts away from the microphone stays audible. It doesn’t squash individual words. On our uneven test file, passages 6 dB apart came out within 2 dB, and the loudness range fell from 17 to 6 LU; the pauses between sentences rose about 3 dB against the voice. For a car or a noisy commute, even out; for headphones at home, louder is usually enough.
Books you can’t boost
Audible’s AAX files, Libby’s and Spotify’s audiobooks are protected: they play only in their own apps, and changing them would mean removing DRM, which this site never does. Turn them up with your phone or car instead. Everything DRM-free works: LibriVox, Libro.fm and Downpour downloads, ripped CDs, your own recordings. Not sure what you have? The audiobook inspector tells you, and the Audible AAX guide explains the protected formats.
Making an audiobook rather than listening to one? Use ACX mastering, which hits ACX’s own levels and format.
Sources
- Loudness normalization on Spotify, Spotify for Artists, checked 23 October 2026
- Audio requirements, Apple Podcasts for Creators, checked 23 October 2026
- EBU R 128: Loudness normalisation and permitted maximum level of audio signals, European Broadcasting Union (v5.0, 2023)
- What are the ACX audio submission requirements?, ACX Help