LUFS explained: how loudness is measured and what to master to

What LUFS measures compared with dBFS

Two files can reach the same highest sample value yet sound different in loudness. That is the distinction between measuring a peak and measuring how loud the material sounds over time.

LUFS stands for loudness units relative to full scale. It is the loudness unit defined by ITU-R BS.1770, a standard for measuring programme loudness. The measurement first applies K-weighting, a filter that approximates how the ear weighs different frequencies.

The meter then averages the filtered signal’s power over time and expresses the result in LUFS. One LU, or loudness unit, represents one decibel of loudness difference. Two files with identical peak levels can differ by 10 LU in loudness.

dBFS means decibels relative to full scale. For sample peaks, it reports the highest sample value rather than loudness over time. A matching peak reading therefore does not mean two masters will sound equally loud.

Integrated, short term and momentary loudness

LUFS, or loudness units relative to full scale, can describe the whole track or a smaller slice of it. The difference is the measurement window.

Integrated loudness is the reading across the whole programme, such as your complete master. It uses gating, a process that excludes silence and very quiet passages from the calculation. A pause therefore does not drag the reading down. Think of integrated loudness as the whole-track view, rather than a reading of one section.

Short-term loudness measures a three-second window. It gives you a closer view of a passage instead of the complete track. Use it to compare sections without treating either section as the overall reading.

Momentary loudness measures a 400-millisecond window. It shows loudness over a shorter slice than short-term measurement. These are three time scales for measuring loudness, not three different units. Check which reading you are looking at before comparing numbers.

What loudness range tells you

A track can move from a quiet verse to a louder chorus without its whole-track reading showing that contrast. Loudness range, or LRA, describes how much short-term loudness varies across a programme. It is measured in loudness units, or LU, with one LU equal to one decibel of loudness difference.

Integrated loudness is the loudness measured across the whole programme, with silence and very quiet passages excluded by gating. It gives you an overall reading. LRA describes the variation within that programme, rather than another version of its overall level.

Two masters can therefore share an integrated reading while differing in loudness range. One might stay close to the same short-term loudness throughout. The other might move between quieter and louder sections. The overall number alone does not tell you which pattern you have.

Why true peak needs a separate measurement

A signal can reach its highest point between two recorded samples. A meter that checks only sample values will not show that higher point. True peak is the estimated maximum level of the reconstructed waveform, rather than just the highest sample value.

The distinction matters because those samples describe a waveform whose peaks can sit above the samples themselves. True peak measurement uses oversampling to estimate what happens between them. It reports the result in dBTP, meaning decibels true peak.

Loudness measurement answers a different question: how loud the material sounds over time. It does not replace a check of the waveform’s highest point. Two files can share the same peak level while differing by 10 loudness units, so neither reading tells the whole story.

What LUFS to master to for streaming

For Spotify, the documented recommendation is about -14 LUFS integrated, with true peak below -1 dBTP. LUFS means loudness units relative to full scale; integrated loudness is the reading across the whole track. True peak estimates reconstructed waveform peaks, expressed in dBTP (decibels true peak).

The AES streaming recommendation, TD1008, suggests -16 to -20 LUFS integrated for streaming services and on-demand distribution, with -1 dBTP true peak. That is broader streaming guidance, not the same recommendation as Spotify’s.

European broadcast has a separate target. EBU R 128 sets -23 LUFS integrated, with a maximum true peak of -1 dBTP. Keep that broadcast specification separate from streaming guidance rather than treating it as another streaming target.

Loudness normalization is playback adjustment toward a similar level. Spotify turns louder masters down and raises quieter ones only as far as their peaks allow. A louder master therefore does not automatically mean louder normalized playback.

Why genre targets are not a mastering standard

Two tracks in the same genre can call for different mastering choices. A genre label does not tell you how much contrast the music should keep between its quieter and louder passages. Treat a genre loudness number as a creative reference, not a documented delivery standard.

The documented recommendations discussed above concern delivery and playback, not a separate target for every genre. They do not establish a loudness requirement for competitiveness. Choose a level that serves the track, rather than treating another producer's number as a rule.

How to measure LUFS on your master

Once you have chosen a version, measure it from start to finish. Use a meter that reports LUFS, loudness units relative to full scale, and reset it before the full track pass.

• Read integrated loudness, the overall loudness across the track, after playback finishes. Its gating excludes silence and very quiet passages so pauses do not pull the reading down.
• Check short-term loudness, measured over three seconds, for individual passages such as a verse or chorus. Momentary loudness uses a 400 millisecond window for briefer changes.
• Note loudness range, or LRA, which describes how much short-term loudness varies across the track. It uses loudness units, with one unit equal to one decibel of loudness.
• Check true peak separately. It estimates the reconstructed waveform's highest level, including peaks between samples, and is expressed in dBTP, decibels true peak.