Put a song from a loud, aggressively mastered album next to a quiet, dynamic one in the same playlist, and on a streaming service they come out at roughly the same volume. Play the same pair from your own files, or from a CD rip, and the loud one buries the quiet one. That difference did not come from the music. It came from the platform deciding, on its own, how loud each file is allowed to sound.
This is loudness normalization, and once you know it is there, a lot of things about modern mastering start to make sense, including why some engineers have quietly stopped chasing maximum loudness at all.
What a meter actually hears
A digital file has a peak level: the loudest single sample, capped at 0 dBFS. For decades that peak was the only number anyone optimized for, because turning a track up against that ceiling was the only way to make it sound competitively loud on the radio or in a shuffled playlist. Push the average level close to the peak with heavy limiting, and the track sounds loud everywhere it plays.
The problem is that peak level and perceived loudness are not the same thing. Two tracks can share the exact same peak and sound completely different in loudness, depending on how much of the track sits near that peak versus how much room it has below it. A meter that only watches the peak cannot tell the difference.
LUFS (Loudness Units Full Scale) is a different kind of measurement. It integrates level over time using a weighting curve tuned to match how the ear perceives loudness, so it produces one number that tracks how loud a piece of audio actually feels, not how close it gets to clipping. This is the number streaming platforms measure, and it is the number they normalize against.
How the turn-down actually happens
The mechanism is simple once you see it. A platform analyzes a track’s integrated LUFS value, either at upload time or on the fly, compares it to a target loudness it has chosen for the whole catalog, and applies a gain offset so playback lands on that target. A track mastered hot gets turned down. A track mastered quiet may get left alone or turned up slightly, usually with a ceiling so it cannot introduce clipping that was not in the original file.
Nothing about the file itself changes. The waveform, the dynamics, the mix, all of it stays exactly as mastered. Only the output gain moves, and it moves by a fixed amount for the whole track, not dynamically like a compressor would.
Why a quiet master can now win
This is where the incentive flips. Under the old peak-based system, mastering a track hot bought you something real: it played louder than the quieter track next to it, everywhere it was heard. Under loudness normalization, that advantage disappears, because the platform just turns the hot master back down to match everything else. You paid for the loudness with squashed dynamics and clipped-sounding peaks, and then the platform took the loudness away anyway.
A dynamic master that was never pushed that hard does not have that problem. It reaches the same target loudness with its transients intact, its quiet passages still quiet, its loud passages still landing with impact instead of arriving pre-flattened. At matched loudness, the ear reads dynamic range as clarity and punch. A track with none left just sounds tiring, even though it is no longer louder than anything around it.
This is also why loudness normalization does not undo bad mastering. If a track’s peaks were clipped during production, turning the whole track down afterward does not restore what was lost. The distortion is baked into the file at a quieter volume, not repaired. Normalization only controls level. It has no opinion on what happened to the waveform before it got there.
The target number moves, and that is fine
Platforms do not all agree on one exact target, and the number a given service uses has changed over time as loudness standards have shifted across broadcast and streaming. A commonly cited target across several major platforms sits somewhere in the neighborhood of minus 14 LUFS integrated, though the precise figure and how strictly it is enforced differs by service and by year.
What matters for a listener is not memorizing that number. It is understanding that a target exists at all, and that it is applied per track, which is why an album can sound perfectly consistent on CD and slightly uneven when streamed, if some tracks were mastered closer to the target than others.
Where ZonaReq sits in this
ZonaReq does not know or care what a streaming app did upstream. It sits at the system audio level, after Spotify, Apple Music, a browser or anything else has already decoded and, if the app applies it, already normalized the track. Whatever level arrives at the output device is what ZonaReq’s equalizer and volume boost work on.
That has one practical consequence worth knowing. Because streaming normalization often turns tracks down rather than up, a lot of what reaches the system is quieter than the file’s own peak would suggest. ZonaReq’s volume boost is built for exactly that gap, and its limiter is there so pushing a normalized-down track back up does not reintroduce the clipping that normalization was designed to route around. It cannot bring back dynamics a track never had. No software downstream of a bad master can do that.
What this means for you
Stop treating inconsistent loudness between tracks as a settings problem to chase. If you are listening on a mainstream streaming service, that leveling is already handled upstream, automatically, per track. If two songs still feel wildly different in impact once they are loudness-matched, that is telling you something about the mastering, not about your volume slider.
It also means loudness is no longer a reason to avoid an equalizer. Boosting a preset or adding a touch of gain will not make a track “win” against a normalization target the way it once could against a raw file, so there is no penalty to shaping tone for how you want it to sound instead of how loud it can be made to appear.