Why Volume Varies Between Songs (and How to Even It Out)

The root cause: the loudness war

Since the late 1990s the music industry has been locked in a strange contest: making albums sound as loud as possible.

The logic made sense at the time. On radio and on shuffling CD players, a louder track feels more "present" and grabs attention. So every label asked for a master louder than its competitors'.

The method is dynamic compression: lifting quiet passages and holding down loud ones, so the whole track sits near maximum.

The consequence is that albums from different eras differ enormously in loudness. A 1985 jazz album and a 2005 rock album can be a dozen decibels apart at the same system volume.

Why ordinary "normalisation" is not enough

The naive approach to evening out volume is to raise every track until its peak touches maximum.

The problem is that peaks do not represent perceived loudness. A track can have one very tall drum hit while the rest sits quietly. Matching by peak makes that track sound even quieter relative to the others.

Human hearing responds to average energy over time, not to the tallest momentary spike.

ReplayGain: measuring the way ears hear

ReplayGain is a standard that analyses a track and calculates how loud it sounds to a person, rather than how tall its peaks are.

The result is stored as a number in the file's metadata — for example "-6.5 dB". That means: to reach the reference loudness, this track needs to come down by 6.5 dB.

The crucial point: ReplayGain does not alter the audio in the file at all. It only writes a note. Players that support it read the note and adjust volume during playback.

That makes the whole process reversible. If you dislike the result, switch it off in the player — the file is exactly as it always was.

Two modes you need to choose between

• Track gain levels each song individually. Right for shuffled playlists drawing from many albums.
• Album gain levels between albums while preserving the volume differences within one album. Right when you listen to albums end to end.

Why album gain matters: some albums are deliberately built around dynamics. A quiet opener that swells in the middle is an artistic decision. Track gain flattens it and erases the intent.

Most players can store both and pick automatically based on how you are listening.

The newer standard: LUFS

More recently the industry moved to a measurement standard called LUFS, which the major streaming services use to normalise their entire catalogues.

This is why volume feels consistent on streaming — they apply automatic normalisation across everything. The side effect is a good one: because excessively loud tracks simply get turned down, the loudness war stopped paying off, and modern masters have begun regaining dynamics.

For a personal collection, ReplayGain and LUFS both work. What matters is picking one and applying it consistently.

What you actually need to do

The work is a one-time pass per file:

• Run a ReplayGain analysis across the whole collection — players like foobar2000 and MusicBee include this built in
• Make sure the ReplayGain option is enabled in your player, otherwise the stored numbers are ignored
• Choose album mode for full-album listening, track mode for mixed playlists

Analysis takes a while on a large collection, but only happens once. After that, every player supporting the standard reads the same notes.

The connection to metadata

ReplayGain values live inside the file's tags, in the same place as title and artist. That means they can be lost if your metadata is damaged or overwritten by a careless tool.

This is one reason cleaning up metadata and levelling volume are best treated as one job — they live in the same place and are lost the same way.

EarTea writes metadata in a way that preserves tags it does not manage, so ReplayGain values you have already computed are not wiped out when titles and artists are tidied.