What is LUFS? Loudness explained

What LUFS actually measures

LUFS stands for Loudness Units relative to Full Scale. It comes from the ITU-R BS.1770 standard and the EBU R128 recommendation built on it. One LU is one decibel of loudness change, so a track at −10 LUFS is 4 dB louder than one at −14 LUFS.

Unlike a peak meter, LUFS is designed to match perceived loudness. The signal is first filtered with K-weighting, which reduces the lowest frequencies and slightly lifts the upper mids, roughly as our hearing does. The weighted signal is then averaged over a time window. That's why a bass-heavy mix can have high peaks but a lower LUFS reading than a bright, dense one.

Momentary, short-term and integrated

The same measurement is shown over three time scales, and each one answers a different question.

Integrated loudness uses gating: very quiet blocks (below −70 LUFS absolute), and blocks more than 10 LU below the running average, are left out. That way long silences and fade-outs don't drag the number down. It also means the integrated reading is only meaningful over the whole song or programme.

Loudness range (LRA)

LRA, measured in LU, describes how much the loudness varies across the programme. It's calculated from the spread of short-term loudness readings (the range between the 10th and 95th percentiles, with gating), so a single loud hit or a quiet intro doesn't dominate it.

A heavily compressed pop master might have a small LRA; a film mix or a classical recording a much larger one. LRA isn't a quality score. It's a description, and some delivery specs give a maximum for it. For a mix engineer it's a useful sanity check: if the chorus and verse are meant to contrast, a very low LRA may mean the mix bus is doing more than you think.

True peak

Digital audio stores samples, but the analogue waveform rebuilt from them can go higher than any single sample. These are inter-sample peaks. A master that never exceeds 0 dBFS on a sample-peak meter can still clip in a converter or after lossy encoding.

A true-peak meter oversamples the signal to estimate those peaks and reports them in dBTP. Most delivery specs set a ceiling between −1 and −2 dBTP. See true peak explained for more on why and how to limit for it.

Normalization versus delivery specs

These are two different situations, and mixing them up causes a lot of confusion.

Full details, tolerances and notes are on the loudness targets page. Targets change over time, so check the service's own documentation before a critical delivery.

What this means while mixing

• Loudness isn't a mix quality. On a normalized service, a louder master plays at the same level as a quieter one. What changes is how much dynamics it has left.
• Listen at the playback level. If your master is −8 LUFS and the service plays at −14, listen to it 6 dB down. That's how it will be compared with everything else in the playlist.
• Watch true peak on loud masters. Heavy limiting and bright content produce more inter-sample peaks, and lossy encoding can add more.
• Measure the whole song. A chorus reading is short-term loudness, not integrated.

Measuring loudness on your iPhone with MixOutpost

MixOutpost's Loudness view shows momentary, short-term and integrated loudness (EBU R128), LRA, true peak and maxima, measured on the phone from the exact audio your Mac sends. Pick a music streaming target, a TV and film delivery spec, or (optionally, in Settings) Apple Podcasts, and the app shows how far the track is from that reference, for example how many dB a service would turn it down.

The measurement happens before Listen as, Dim, Mute, the diagnostic EQ and phone volume, so you can listen on the phone speaker while the reading stays accurate. Reset, play from the top, and you have an integrated reading without bouncing. The loudness meter passes the EBU test cases in automated tests.

In the proposed free and Pro split, momentary and short-term loudness are free, and the integrated, LRA, true peak and target readings are Pro.