How to Normalize Audiobook and Podcast Audio to a Consistent Volume

Chapter three is whisper-quiet, chapter four blows your earbuds out; last week’s episode needs the volume up, this week’s needs it down. Spoken audio has the same disease as music libraries, but it answers to two different rulebooks: audiobooks inherited an 89 dB tradition from the ReplayGain world, while podcasts standardized on LUFS. MP3Gain genuinely fits one of those jobs and only half of the other. Here is the honest map, and the workflow for each.

Audiobooks: the 89 dB tradition is MP3Gain’s home turf

The world’s largest volunteer audiobook project, LibriVox, has asked narrators for the same thing for years: an overall volume of around 89 dB — its wiki gives an acceptable band of roughly 87–91 dB, ideally 89 — so that consecutive chapters and different narrators sit at one level. LibriVox even ships a small program, Checker, that measures an MP3’s volume and technical specifications so contributors can verify a file before submitting it. That 89 dB figure is ReplayGain thinking: the same reference MP3Gain uses as its default target. Which means the tool already on this site speaks the audiobook world’s language natively, no translation layer required.

The chapter workflow. Collect the finished chapter MP3s in one folder. Add the folder to MP3Gain and run Track Analysis: the per-file values immediately expose the outlier chapters. Apply Track Gain at the default 89.0 dB target with the clipping guard on. Spot-check the loudest and quietest chapters by ear, then run the files through Checker (for LibriVox submissions) or simply play chapter boundaries back-to-back. The whole pass is lossless and reversible — if a narrator re-records one chapter, re-run just that file to the same target and the set matches again. For libraries of hundreds of files, the backup-first routine in batch normalizing a large MP3 library applies unchanged.

The boundary: gain cannot fix dynamics inside a chapter

MP3Gain applies one gain value per file (or one shared value per album). That fixes file-to-file consistency perfectly — and does nothing about a chapter that whispers its dialogue and shouts its narration, because that unevenness lives inside the file. Evening out within-file dynamics is compression’s job, done in an audio editor before the leveling pass, and no normalizer of any brand changes that division of labour; LibriVox’s own guidance makes the same point, reaching for compression when plain amplification is not enough. Likewise, MP3Gain will not remove hiss, repair clipping, or rescue a bad recording — it preserves the file faithfully at a better level. Sequence your fixes honestly: edit and compress first, level last.

Podcasts: a different rulebook, measured in LUFS

Podcasting standardized on the broadcast measurement instead. The industry convention — the target Apple Podcasts has long published — is about −16 LUFS integrated for stereo episodes and −19 LUFS for mono, with true peaks kept around −1 dBTP. Those are BS.1770 loudness numbers (the full ReplayGain-versus-LUFS comparison is in ReplayGain vs LUFS), and they are measured with a loudness meter, typically at the final master in your editor. That is the step that makes episodes match each other and sit correctly next to every other show in a listener’s app — some platforms also normalize playback, but delivering at the conventional target is what keeps your show consistent everywhere, including the apps that do not.

Where MP3Gain fits podcasting — and where it stops. For a finished stereo master that must hit −16 LUFS, MP3Gain is the wrong instrument: it neither measures LUFS nor touches true peak, so it cannot certify the number a platform expects. Its legitimate podcast role is narrower and practical: a back catalog of MP3 episodes that are uniformly out of line with each other — an early season mastered timidly, a later one mastered hot. A Track Gain pass at one shared reference makes old episodes consistent with each other in one reversible batch, which is a real improvement for binge listeners; then verify a sample episode with a LUFS meter and, if it must meet the −16 LUFS convention exactly, finish the job in an editor’s loudness normalization. Use each tool for the measurement it actually makes.

For narrators and hosts: leveling is the last step, not the first

Everything above assumes finished files. If you are the person recording them, the cheapest loudness fix happens before any software runs: record every chapter or episode at the same input level, in the same room position, with the same microphone distance. A narrator who sits in the same place and watches the same meter range produces chapters that barely need leveling; a host who records one episode at a desk and the next in a car produces files no normalizer can fully reconcile, because the room tone and dynamics changed along with the volume. Treat recording consistency as the foundation, editing and compression as the shaping, and the final gain pass — MP3Gain for audiobook-style targets, a LUFS meter for podcast masters — as the inspection stamp at the end. Skipping to the stamp is how volumes end up technically matched and obviously different.

“Can I just run my podcast through MP3Gain and call it −16 LUFS?” No — and the reason is worth keeping straight, because the two numbers are answers to different measurements. MP3Gain will make your episodes consistent with each other at a ReplayGain reference; −16 LUFS is a BS.1770 integrated loudness value with a true-peak condition attached. Episodes leveled by MP3Gain will often land in a plausible neighbourhood, but “plausible neighbourhood” is not the deliverable a platform’s spec describes, and the honest final step for a podcast master is a loudness measurement in a tool that reports LUFS and true peak directly. Consistency first, certification second, each with its own instrument.

One-page summary

Spoken word rewards consistency more than music does — a listener in a car, on a run, or falling asleep to a chapter cannot keep reaching for the volume, and will blame the book or the show, never their own playback chain. Match the rulebook to the job, and both jobs are an afternoon’s work. One practical closing note for both crafts: keep an unleveled archive copy of every finished file. Leveling with MP3Gain is reversible from its own tag data, and a LUFS master can always be re-exported — but an untouched archive makes every future standard change a fresh start instead of an archaeology project, and standards in spoken audio have changed before.