You may play two songs at the same volume setting and find that one sounds much louder than the other. The same thing can happen when moving between videos, podcasts, or other audio. Reaching for the volume control each time is inconvenient, so many playback systems offer some form of loudness normalization.
Loudness normalization tries to make different pieces of audio start from a more consistent perceived loudness. It does not flatten every quiet and loud moment into one level, and it does not guarantee that every recording will sound identical in volume.
Understanding that distinction explains both why normalization is useful and why you can still hear meaningful volume differences while it is enabled.
The volume control is only one part of what you hear
Your device’s volume setting controls how strongly the playback system sends audio to the speakers or headphones. But the audio itself also has a recorded level and a pattern of quiet and loud sounds.
Imagine playing a softly mastered acoustic track followed by a densely mastered pop track. If you leave the device volume unchanged, the second track may sound noticeably louder because the two recordings were prepared differently.
This is why a 50% volume setting does not correspond to one fixed perceived loudness for all content. The setting controls playback gain, while the content determines what signal is being amplified.
Normalization addresses the difference between pieces of content rather than simply choosing a new position for your volume slider.
Loudness is not the same as the highest peak
A digital audio signal has peaks: brief moments where its level reaches a high value. Measuring only the highest peak is not enough to predict how loud an entire song or programme will seem.
Consider two recordings with the same highest peak. One might contain a single strong drum hit surrounded by relatively quiet music. The other might remain close to its peak level for much of the track. Even though their maximum peaks are similar, the second recording can sound louder overall.
Loudness measurement therefore considers audio over time rather than looking only at its single highest sample. Modern loudness measurement methods also account for the fact that human hearing is not equally sensitive to every frequency.
One widely used technical basis for this kind of measurement is ITU-R BS.1770, which defines algorithms for measuring programme loudness and true-peak level. Different services can build their own playback policies around such measurements, so the exact target and behaviour are not universal.
Normalization usually adjusts playback gain
The basic idea is straightforward. A system estimates the overall loudness of a track or programme, compares that measurement with a chosen reference, and applies an appropriate gain adjustment during playback.
If a track measures louder than the reference, the player can turn it down. If a track measures quieter, the system may be able to turn it up, although available headroom and the service’s rules can limit how much upward adjustment is practical.
Suppose, purely as an example, that one track is measured several loudness units above a service’s reference. The player can reduce its playback gain so it begins at a level closer to other normalized tracks. The original audio file does not necessarily have to be rewritten; the adjustment can be applied by the playback system.
The important point is that normalization is based on measured content loudness, not on whether your device volume slider happens to be at 20%, 50%, or 80%.
It does not make every second equally loud
A well-normalized recording can still move from a quiet verse to a powerful chorus. A film can still have quiet dialogue followed by a louder musical section. Normalization does not inherently remove those internal contrasts.
This difference is easier to understand by separating overall loudness from dynamic range. Overall loudness describes the general level of a piece of audio across time. Dynamic range describes variation between quieter and louder parts.
If a player turns an entire track down by the same amount, its internal relationships remain. A drum hit that was louder than a guitar before the adjustment is still louder afterward. The whole signal has simply moved to a lower playback level.
Processes such as dynamic-range compression can deliberately reduce the difference between quieter and louder passages. That is a different operation from ordinary loudness normalization, although a particular product or listening mode may combine several kinds of audio processing.
Why a quiet recording cannot always just be turned up
Turning down a loud track is relatively straightforward because it creates more room below the system’s maximum signal level. Turning up a quiet track can be more complicated.
A recording may already contain peaks close to the highest level that its digital representation or playback chain can accommodate cleanly. Raising the whole signal enough to reach a loudness target could push those peaks beyond the available headroom.
A service then has choices. It can apply less upward gain than the loudness difference suggests, use additional processing such as limiting, or follow some other platform-specific policy. Different services and applications make different decisions.
This is one reason normalization should not be understood as a promise that every item will reach exactly the same audible level.
Perceived loudness still varies after normalization
A loudness measurement is designed to correspond more closely to human perception than a simple peak meter, but it cannot predict every listening situation perfectly.
Several factors still matter:
- the frequency balance of the recording;
- how much its level changes over time;
- the speakers or headphones being used;
- background noise in the room;
- the listener’s hearing and preferences;
- any equalizer, sound enhancement, or other processing in the playback chain.
A speech recording and a bass-heavy music track can therefore feel different even when a system has normalized their measured loudness. The goal is usually to reduce large, distracting differences, not to make unrelated material perceptually indistinguishable.
Album listening creates a special trade-off
Individual-track normalization can be helpful when a playlist mixes music from many artists and releases. Each track can be brought closer to a common reference, reducing the need for manual volume changes.
Albums can present a different problem. An artist may intentionally make one track quieter than the next, or design several tracks to flow together with specific relative levels. Normalizing each track independently can alter those intended relationships.
Some playback systems therefore support album-aware behaviour, where the relative loudness differences within an album are preserved while the album as a whole receives a gain adjustment. Whether this is available, and when it is used, depends on the application or service.
The distinction matters most when you care about the level relationships chosen across a complete release. For a mixed playlist, track-by-track consistency may be more useful.
Normalization does not repair poor audio
Loudness normalization changes level. It cannot restore detail that was lost in a low-quality recording, remove distortion that is already present, improve a noisy microphone, or recreate dynamic range that was removed during production.
It also does not mean that a louder master automatically becomes equivalent to a more dynamic one. If two recordings were mixed and mastered differently, lowering one to a common playback loudness does not undo those production choices.
This is useful when comparing audio quality. A recording that initially seems more impressive simply because it is louder may lose that immediate advantage when both recordings are heard at similar loudness. Differences in clarity, balance, distortion, and dynamics can then be easier to judge on their own merits.
When to use loudness normalization
Normalization is especially useful when you frequently move between unrelated tracks, videos, or programmes and find yourself adjusting the volume to compensate for large changes in level.
If your player offers the feature, try it with the kind of listening you actually do. Settings and names vary between applications, and some services provide more than one normalization mode. There is no need to assume that enabling it will improve every listening preference.
For critical listening to an album, you may prefer a mode that preserves album-level relationships if the software provides one. If you are troubleshooting audio that sounds unexpectedly quiet, also check ordinary causes such as the device volume, application volume, output device, equalizer settings, and other sound processing. Normalization is only one part of the playback chain.
The practical takeaway
Loudness normalization is best understood as an automatic gain adjustment based on measured audio loudness. It helps reduce large level jumps between pieces of content while leaving the listener’s main volume control available for choosing a comfortable overall listening level.
It is not the same as peak matching, and it does not necessarily compress the quiet and loud moments inside a recording. Because playback policies and listening conditions vary, perfect consistency is neither guaranteed nor usually the goal.
The useful mental model is simple: your volume control sets how loud you want to listen; normalization helps different recordings arrive at that control with more comparable overall levels.