What Is ReplayGain? The Standard Behind MP3Gain

ReplayGain is a method for measuring how loud a piece of audio sounds to a human listener — rather than how high its peaks are — and expressing that measurement as a single gain adjustment relative to a reference loudness of 89 dB SPL. It was proposed by audio engineer David Robinson in 2001, in a document titled “Replay Gain — A Proposed Specification.” It is the algorithm inside MP3Gain and foobar2000, the reason every ReplayGain tool targets the same 89 dB, and still the standard way to make a personal music library play at one consistent volume.

The problem it was built to solve

By 2001, anyone with a folder of MP3s from different CDs had the same daily irritation: every album played at a different volume, because every mastering engineer had chosen a different level — and the trend, the “loudness war,” was toward ever hotter masters. The obvious fix of the time, peak normalization, barely helped: it equalized the loudest instant of each track while leaving perceived loudness all over the place, since a single drum hit says nothing about how loud a song feels. Robinson’s proposal started from psychoacoustics instead: measure the loudness the way ears experience it, and normalize that.

How the measurement works

A ReplayGain analysis runs through three ideas. First, the audio passes through an equal-loudness filter that mimics human hearing — we are most sensitive to midrange frequencies and relatively deaf to deep bass at low levels, so the measurement weights frequencies the way an ear does. Second, the filtered signal’s energy (RMS level) is sampled in short blocks across the whole track. Third, instead of averaging those blocks — which silence and quiet passages would drag down — the specification takes a value near the top of the distribution (the 95th percentile), representing the level the track sustains when it is actually playing. That figure is compared against the reference, and the difference is the ReplayGain adjustment: “this track needs −6.3 dB.”

The specification defines two modes, which MP3Gain users know well: Track Gain (originally “Radio Gain”), where every song is brought to the reference independently, and Album Gain (“Audiophile Gain”), where one shared adjustment preserves an album’s internal dynamics. Their differences and uses are covered in Track Gain vs Album Gain.

Where the 89 dB came from

ReplayGain needed a fixed point, and Robinson borrowed one from the film industry. The SMPTE calibration standard for cinema sound says a pink-noise test signal at −20 dB relative to full scale should play back at 83 dB SPL (sound pressure level) in the theatre. The original proposal worked from that 83 dB reference — but implementations quickly settled on playing the reference 6 dB hotter, equivalent to pink noise at −14 dBFS, which on the same calibrated scale is 89 dB SPL. Robinson endorsed the change, and 89 dB became the de facto reference for every ReplayGain tool — it is the “Target Normal Volume” you see in MP3Gain, whose full story is in the 89 dB explainer. The point of any reference is not the number itself; it is that everyone uses the same one, so an analysis made in one program means the same thing in another.

A measurement, not a change: tags and players

ReplayGain itself only defines the measurement and the values. In its most common form, those values are written into a file’s metadata tags (for MP3s, typically APEv2 or ID3v2 tags), and a supporting player applies the adjustment during playback — the audio data is never modified. foobar2000 is the best-known player that both scans and applies ReplayGain this way, and FLAC and Vorbis files carry ReplayGain fields as part of their format conventions.

MP3Gain is the famous exception, and the reason it exists as a separate program: instead of relying on players to honor tags, it applies the ReplayGain adjustment directly to the MP3 frames’ gain values, so the correction works in every player and device ever made — at the price of working only on MP3s. That mechanism is detailed in Is MP3Gain Lossless? and the MP3Gain overview.

Support in practice: broad, but not universal

On the desktop, ReplayGain support is deep: foobar2000 scans and applies it natively, as do MusicBee, MediaMonkey, and most Linux players, and tag editors can write the values without playing a note. The picture thins out as you move away from computers. Many hardware players, car head units, and smart speakers ignore ReplayGain tags completely, and Apple’s ecosystem uses its own Sound Check system instead — similar in spirit (a per-track loudness value stored in metadata) but not interchangeable with ReplayGain values. This patchiness is not a footnote to the ReplayGain story; it is the direct reason MP3Gain’s file-level variant matters. A gain change written into the MP3 frames needs no support from anyone, which is why a tag-based standard ended up with a tag-free flagship tool.

ReplayGain today

ReplayGain was a proposal, never an ISO standard — it spread because it worked and because the specification was public. Two decades on, it remains the native loudness system of the file-based music world: players, taggers, and formats still support it, while the streaming and broadcast industries standardized on the younger, absolute LUFS scale (ITU-R BS.1770). A community revision, ReplayGain 2.0, later adopted BS.1770-style measurement under the ReplayGain name, bringing the two systems closer; the three-way comparison explains how they relate. Its values also became a shared vocabulary: when experienced users say a master “measures about 97 dB,” everyone in the conversation knows which ruler produced the number and what it implies. For the original job — a hard drive full of music that should play at one volume — the 2001 design, and MP3Gain with it, never needed replacing.