ReplayGain vs LUFS vs Peak Normalization: Three Ways to Measure Loudness

Peak normalization, ReplayGain, and LUFS answer three different questions. Peak normalization asks: “How close is the loudest instant to the maximum the format can store?” ReplayGain asks: “How loud does this track sound, compared to a reference of 89 dB SPL?” LUFS — the scale defined by the ITU-R BS.1770 standard — asks: “How loud does this sound, in absolute terms, integrated over the whole track?” Streaming services normalize with LUFS; MP3Gain uses ReplayGain, because it was built for personal music libraries in 2001, a decade before BS.1770 loudness became the broadcast and streaming norm.

The three systems side by side

Peak normalization: the loudest instant wins

Peak normalization is the oldest and crudest method: find the highest sample in the file, then scale the whole file so that peak touches a set level, usually 0 dBFS. Every audio editor’s “Normalize” button historically did exactly this. Its flaw is fundamental — a peak is an instant, not an impression. A track with one loud drum hit and an otherwise quiet performance peak-normalizes to the same value as a wall of compressed sound, yet the second sounds dramatically louder. Two songs can both sit at 0 dBFS peak and differ by 10 dB or more in perceived loudness. Peak levels still matter as a safety check (clipping is a peak problem), but as a loudness measure, peak normalization answers the wrong question.

ReplayGain: loudness relative to 89 dB

ReplayGain measures how the audio actually sounds: the signal is filtered to mimic the ear’s frequency sensitivity, its energy is averaged in short blocks, and a high percentile of those blocks — rather than a single peak — becomes the track’s loudness. The result is expressed as a gain adjustment relative to the shared 89 dB reference. That relative design is why ReplayGain values travel so well between tools: MP3Gain, foobar2000, and supporting players all measure from the same fixed point, so a track analyzed once is understood everywhere. The full story is in What Is ReplayGain?

LUFS: the absolute scale the industry moved to

LUFS (loudness units relative to full scale; the identical scale is also written LKFS) comes from ITU-R BS.1770, a standard built for broadcasters who needed different networks and countries to agree on loudness. It uses K-weighting — a defined frequency weighting — plus gating that excludes silence and quiet passages from the integrated measurement, and it reports an absolute number: a track is −14 LUFS, full stop, no reference point to know. Regulation and platforms adopted it: European broadcast targets −23 LUFS (EBU R128), and streaming services normalize playback to their own targets, commonly around −14 LUFS for Spotify and YouTube and −16 LUFS for Apple Music.

A later development, ReplayGain 2.0, adopted BS.1770-style measurement inside the ReplayGain framework, and a documented conversion between the two scales exists — the two worlds are cousins now, not strangers. But classic MP3Gain implements the original 2001 algorithm, and for balancing a music library that remains entirely fit for purpose — for the MP3 files that dominate older collections, its readings are still the practical common language between tools.

Which one should you care about?

• Making your own library play evenly: ReplayGain, via MP3Gain. It is the only one of the three designed for exactly this, and it works at the file level, in every player.
• Publishing music or video to streaming platforms: LUFS. Master to the platform’s target (around −14 LUFS for music streaming) so their normalization doesn’t make the decision for you.
• Checking technical headroom: peak levels. A peak readout tells you how close you are to clipping — a safety gauge, never a loudness target.
• Podcasts and spoken word for distribution: LUFS again — podcast platforms publish loudness targets (commonly −16 LUFS for stereo podcasts), and meeting them keeps episodes consistent across apps in a way no peak setting can.

The mistake to avoid is mixing the scales: an MP3Gain reading of “91.5 dB” and a meter reading of “−11 LUFS” describe the same song through different rulers, and neither number can be typed into the other system.

Frequently asked questions

Is LUFS better than ReplayGain?

For broadcasters and streaming platforms, LUFS is the better tool: it is an absolute, regulated standard with gating that handles mixed content like speech and adverts. For leveling a personal MP3 library, ReplayGain is equally accurate in practice and is what the available tools implement. “Better” depends on the job.

Why does MP3Gain show dB instead of LUFS?

Because MP3Gain implements the original ReplayGain specification from 2001, which expresses results as a gain relative to the 89 dB SPL reference. BS.1770 and LUFS did not exist yet. The program’s design predates the scale, and its results remain compatible with every other ReplayGain tool.

Can I convert a ReplayGain value to LUFS?

Approximately. ReplayGain 2.0, which adopted BS.1770 measurement, defines a bridge between the scales (roughly, an integrated loudness of −18 LUFS corresponds to a ReplayGain value of 0 dB). Classic MP3Gain’s 1.0 analysis can differ by a dB or so on some material, so treat any conversion as an estimate, not an exact translation.