Glossary
The loudness words, and the documents that define them.
LUFS, dBTP, LRA, PLR, K-weighting, true peak, the gates, the penalty. Twenty-six terms, each one a short definition and, where one exists, a link to the document that defines it, so a number on this site can be traced to where it came from.
- LUFS
-
Loudness Units relative to Full Scale. One number for how loud a signal is, measured the way ITU-R BS.1770 describes: the signal is K-weighted, split into blocks, the quiet blocks are gated out, and the rest is averaged. Every published streaming target is written in LUFS. Music in practice reads below 0 LUFS, so the numbers are negative and −9 is louder than −14.
Defined by ITU-R BS.1770
- LU
-
Loudness Unit. A difference between two loudness readings; 1 LU is the same size as 1 dB. A track at −9.4 LUFS is 4.6 LU louder than a −14 target. LUFS is a level, LU is a distance.
Defined by ITU-R BS.1770
- LKFS
-
Loudness, K-weighted, relative to Full Scale. The same measurement as LUFS under the name the ITU and the US broadcast recommended practice ATSC A/85 use; the EBU chose LUFS. A reading in LKFS and a reading in LUFS are directly comparable.
Defined by ITU-R BS.1770
- Integrated loudness
-
The loudness of the whole programme, from start to finish, with the gates applied. When a platform or a delivery spec says LUFS with no qualifier, it means integrated. The other two windows exist for watching a meter while you work.
Defined by ITU-R BS.1770
- Momentary and short-term loudness
-
Momentary loudness is measured over a 400 ms window and short-term over 3 s, both K-weighted and both ungated. EBU Tech 3341 defines the two windows and how a compliant meter displays them.
Defined by EBU Tech 3341
- K-weighting
-
The two-stage filter applied before loudness is summed: a high-frequency shelf that lifts the treble, then a high-pass that reduces the lowest bass. It exists because a peak meter counts a sub-bass note and a hi-hat equally and a listener does not. It is why a bass-heavy mix and a bright one can peak at the same level and read differently.
Defined by ITU-R BS.1770
- Gating
-
Two thresholds that decide which 400 ms blocks count toward the integrated reading. The absolute gate discards blocks below −70 LUFS, so silence is ignored. The relative gate then discards blocks more than 10 LU below the loudness of the blocks that passed the absolute gate, so a long quiet intro does not drag a song’s number down.
Defined by ITU-R BS.1770
- Loudness range (LRA)
-
How far the loud parts of a track sit from the quiet parts, in LU. EBU Tech 3342 defines it from the distribution of short-term loudness: after a −70 LUFS absolute gate, a relative gate 20 LU below the absolute-gated loudness removes silence and background noise, and the range is the distance between the 10th and 95th percentiles, the lower of which keeps a fade-out from dominating. Heavy limiting produces low numbers; there is no correct value.
Defined by EBU Tech 3342
- dBFS
-
Decibels relative to full scale, the largest value a digital file can hold. 0 dBFS is the ceiling of the format, and in a fixed-point file every sample sits at or below it. A peak meter reads dBFS. It says nothing about loudness: two files can peak at the same dBFS and be 6 dB apart in how loud they sound.
A general audio term; no single standard defines it
- Sample peak
-
The highest single sample value in a file, in dBFS. It is what a peak meter in a DAW shows and what most clip indicators watch. It misses the level the waveform reaches between samples, which is why true peak exists.
Defined by ITU-R BS.1770, Annex 2
- True peak (dBTP)
-
The highest level the reconstructed waveform reaches, including the points between samples. A converter or a lossy encoder rebuilds a continuous signal from the samples, and that signal can rise above every sample in the file. BS.1770 measures it by oversampling, four times for a 48 kHz signal, and reading the peak of the result. A file whose highest sample reads 0.0 dBFS can reach +7.1 dBTP, measured on the checker’s own clipped-noise test file.
Defined by ITU-R BS.1770, Annex 2
- Inter-sample peak
-
The part of the true peak that a sample-peak meter cannot see: the level between two adjacent samples when the waveform curves above both. Lossy encoders add more of it on decode. It is the reason delivery specs pair a loudness target with a true-peak ceiling below zero.
Defined by AES TD1008
- True-peak ceiling
-
The highest true peak a limiter is allowed to let through. New uploads here use −0.1 dBTP. Older presets also use −0.3 and −0.5, −1.0 (the general recommendation) and −2.0 (what Spotify and SoundCloud ask for above −14 LUFS). The ceiling is never traded for loudness: when a target is out of reach under it, the master is delivered short and the shortfall is stated.
Defined by AES TD1008
- Headroom
-
The distance between a signal’s peak and a limit, usually 0 dBFS or a delivery ceiling. A mix bounced with a few dB of headroom gives a mastering chain room to work; a mix already peaking at 0 has spent it.
Defined by AES TD1008
- Crest factor
-
The ratio of a signal’s peak level to its average level, in dB. Drums have a high crest factor, a sustained pad a low one, and limiting lowers it. It is the property a limiter spends to buy loudness.
A general audio term; no single standard defines it
- Peak-to-loudness ratio (PLR)
-
True peak minus integrated loudness. A master at −1 dBTP and −14 LUFS has a PLR of 13 dB; the same peak at −8 LUFS has 7. It states how much of the available range the master kept, and it does not change when a platform turns a track down: the level drops, the ratio stays.
Defined by AES TD1008
- Limiter
-
A gain reducer that holds peaks under a ceiling, working fast enough to catch every one. A true-peak limiter watches the oversampled signal so the ceiling holds after reconstruction. In this chain it is the only limiter a master goes through, and on the three louder choices a clipper sits ahead of it; the loudest compliant gain is found by search rather than guessed.
Defined by How it works
- Clipper
-
A stage that shaves the top off a peak instead of turning the whole signal down around it. A soft clipper rounds the cut rather than squaring it. Ahead of a limiter it buys back the headroom the limiter would otherwise spend on a few transients, which is why the chain reaches its targets on material a limiter alone left short.
Defined by The mastering chain
- Loudness normalisation
-
Playback gain a platform applies so tracks sit near a target level. It changes the level a listener hears, never the file. Spotify: “We adjust tracks to −14 dB LUFS, according to the ITU 1770 standard.” It is not universal: Spotify documents it as off on its web player and third-party devices, and five of the largest platforms publish no target for a stereo track.
Defined by Spotify for Artists
- Loudness penalty
-
How far a platform turns a track down: the track’s integrated loudness minus the platform’s published target, when the track is louder. A −9 LUFS master on a −14 platform is played 5 LU quieter. Only platforms that publish a figure have a penalty to calculate.
Defined by The loudness penalty calculator
- Level-matched A/B
-
Two versions played at the same loudness so the comparison focuses on tone and dynamics rather than volume. Bounce Sessions (staging) previews intentionally use the original and mastered levels so you hear the delivered loudness change.
Defined by How it works
- Reference master
-
A finished record the chain measures your mix against to derive its tonal correction. Here the reference is one real master chosen by a mastering engineer; the match corrects spectral balance and width toward it and leaves the arrangement alone.
Defined by The mastering chain
- Dither
-
Low-level noise added when a file’s bit depth is reduced, so that rounding errors become a steady hiss below hearing instead of distortion that follows the music. It belongs on every reduction in bit depth, including a 24-bit export from a floating-point session, where it is inaudible; a 32-bit float file needs none.
A general audio term; no single standard defines it
- ITU-R BS.1770
-
The international recommendation that defines how loudness and true peak are measured: K-weighting, the block and gate rules, and the oversampled true-peak meter. The current revision is BS.1770-5 (November 2023); the stereo measurement is unchanged from revision 4, which the engine and the checker implement. The checker is compared against ffmpeg’s implementation of it.
Defined by ITU-R BS.1770
- EBU R 128
-
The European Broadcasting Union’s loudness recommendation for broadcast: programmes normalised to −23 LUFS with a maximum true peak of −1 dBTP, measured as BS.1770 describes. Its companion documents Tech 3341 and 3342 define the meter modes and loudness range. It is a broadcast recommendation, revised in 2023; streaming platforms sit well above its target.
Defined by EBU R 128
- AES TD1008
-
An Audio Engineering Society technical document for distributors of streams and on-demand audio: −16 LUFS when a service normalises track by track, −14 LUFS for the loudest track of an album when it normalises by album, and a maximum true peak of −1 dBTP at the input of a lossy encoder, lower as bitrate falls. It says it does not make recommendations for content production, and it is often quoted as if a platform had published it.
Defined by AES TD1008
Where the numbers live
Three documents define nearly all of it.
ITU-R BS.1770 defines how loudness and true peak are measured. EBU R 128 and its Tech 3341 and 3342 papers define the meter modes and loudness range. AES TD1008 recommends distribution loudness to streaming services. What each platform actually publishes is a shorter list, and it is on its own page.
Read the numbers on your own track.
The free preview measures your mix, masters it to the target you pick, and measures the file that comes back. Both sets of readings, labelled.