FLACCID
All posts
Guides May 5, 2026 · 6 min read

ReplayGain Explained: Consistent Volume Without Ruining Dynamics

ReplayGain evens out volume across your library using metadata tags: no re-encoding, fully reversible, and your music's dynamics stay intact.

You know the move. A gently mastered folk record from 1975 fades out, a 2019 remaster slams in, and you lunge for the volume before your ears (or your neighbours) pay for it. Ten minutes later a quiet jazz side comes on and you're cranking it back up. ReplayGain exists to end exactly this ritual, and it does it in the most polite way imaginable: without changing a single sample of your audio.

That last part is the bit people miss, so let's start there.

What ReplayGain actually does

ReplayGain is a two-step idea: measure once, adjust at playback.

First, a scanner analyses each file and estimates how loud it actually sounds to a human, not the peak level, which tells you almost nothing, but the perceived loudness over the whole track. It compares that to a standard reference level and works out the difference: this track is 7.2 dB louder than the reference, that one is 3.4 dB quieter.

Then it writes the result into the file's metadata, tags like REPLAYGAIN_TRACK_GAIN and REPLAYGAIN_ALBUM_GAIN, sitting right alongside the artist and album fields. That's it. That's the whole intervention.

At playback, a ReplayGain-aware player reads the tag and nudges its own volume by that amount. Loud files get turned down, quiet files get turned up, and everything lands at roughly the same perceived level. It's an invisible hand on the volume knob, doing the exact adjustment you'd have made yourself, just faster and more accurately.

ReplayGain never edits your audio. The samples in the file are bit-for-bit identical after scanning. The only thing added is a few lines of text metadata.

This is why it's completely reversible. Don't like it? Turn it off in the player, or delete the tags. There's no re-encoding, no generational loss, no "normalized" copy of your library eating a second terabyte. Compare that to destructively normalizing files in an audio editor (which rewrites the audio and can never be perfectly undone) and the elegance of the design starts to show.

Track gain vs album gain

Scanners typically write two values per file, and choosing between them is the one real decision ReplayGain asks of you.

  • Track gain brings every individual track to the reference loudness. Perfect for shuffle and playlists, where a whispered ballad might follow a metal track from a different century.
  • Album gain applies one shared adjustment to the whole album, calculated from the album's overall loudness. The album gets louder or quieter as a unit, but the relationships inside it are untouched: the quiet interlude stays exactly as far below the closer as the mastering engineer placed it.

That distinction matters more than it sounds. Albums are mixed with internal contrast on purpose. If you track-normalize a record where the hushed intro is meant to be hushed, you've flattened a deliberate artistic choice. So the rule of thumb: album gain when you listen to albums, track gain when you shuffle. Good players let you pick, and since the tags store both, you never have to commit.

Album gain has a quieter benefit too: on a continuous record, a per-track gain change at a seam is audible as a small step in volume. Album gain applies one value across the whole thing, which keeps gapless playback genuinely seamless.

There's also a pair of peak tags (REPLAYGAIN_TRACK_PEAK and friends). If a quiet track needs a positive gain, the adjustment could theoretically push its loudest moments past full scale; the peak value lets the player see that coming and prevent clipping. You'll rarely think about it, but it's why the system doesn't distort.

Where the numbers come from: the R128 and LUFS lineage

The original ReplayGain proposal dates back to 2001: David Robinson's spec used an equal-loudness filter and some clever statistics to approximate how loud a track feels, referenced to a target of about 89 dB SPL. It was genuinely ahead of its time: hobbyist software solving a problem the broadcast industry hadn't formalized yet.

The industry eventually caught up, and did it rigorously. The ITU's BS.1770 loudness algorithm and the EBU's R128 recommendation defined loudness measurement in LUFS (Loudness Units relative to Full Scale), and it became the standard for broadcast and, later, streaming. Modern ReplayGain scanners (the so-called ReplayGain 2.0 implementations you'll find in tools like foobar2000 and loudgain) use the R128 measurement under the hood, with a reference of −18 LUFS, while still writing the same familiar tags. Old idea, better ruler.

Here's the quiet punchline: every major streaming service now does loudness normalization by default. They just do it server-side, with their own targets, on their own terms. ReplayGain is the same idea for your own library, except you choose the settings, you can turn it off, and the analysis lives in files you control, tagged like any other metadata and travelling with your collection wherever it goes.

The opposite of the loudness war

If you know the history of the loudness war, ReplayGain might sound suspiciously like more of the same, more machinery messing with volume. It's precisely the reverse.

The loudness war happened because louder sounds better in a quick A/B comparison, so masters were compressed and limited to win a contest measured in the first three seconds. The cost was dynamics: the difference between a whisper and a roar got crushed into a uniform, fatiguing wall.

Loudness normalization takes away the prize. If every track plays back at the same perceived level anyway, a crushed master gains nothing: it just gets turned down, and all it has left is its squashed dynamics. Meanwhile a dynamic master, turned up to the same level, keeps its punch, its contrast, its sense of air. Under normalization, dynamics stop being a competitive liability and become an audible advantage again.

One honest caveat: ReplayGain can't restore dynamics that were destroyed in mastering. A brickwalled album is still brickwalled at any volume. What it fixes is your listening session, and what it removes, at scale, is the incentive that caused the damage in the first place.

Getting it into your library

The workflow is short: run a scanner over your collection (foobar2000 on Windows, loudgain cross-platform, and most serious library tools have it built in), choose album mode, and let it write tags. A big library takes a while to analyse (it's reading every sample) but it's a one-time job, and rescanning is only needed for new additions.

From then on, any ReplayGain-aware player just works. Albums keep their internal shape, your volume knob stops being a nervous tic, and nothing about your files has changed except a few lines of text you can delete on a whim.

That's the takeaway worth keeping: consistent volume and preserved dynamics aren't in tension. The loudness war made them enemies; ReplayGain makes them the same feature. Your library can be both even and alive: it just needs to be measured instead of mangled.

Curious how your collection behaves? Open the player, load an old album and a modern remaster back to back, and hear exactly what an even playing field sounds like.

try it yourself

Play your own lossless files, free

FLACCID plays your local FLAC, ALAC, WAV and more right in the browser: no upload, no account, no streaming.

Launch the Player