Momentary, Short-Term, and Integrated LUFS: What Each Reading Means

A loudness meter reports a single quantity — loudness in LUFS (Loudness Units relative to Full Scale) — over several different time windows. The windows are why one meter can show three numbers at once, and why they rarely agree.

All of them come from the same measurement procedure, defined in ITU-R BS.1770: the signal is filtered, squared, averaged over time, weighted per channel, summed, and expressed on a logarithmic scale. Change only the averaging time, and you get a different reading of the same audio.

The filtering step (K-weighting)

Before anything is averaged, each channel passes through the two-stage frequency weighting BS.1770 calls K-weighting: a high-frequency shelf and a low-frequency roll-off. The effect is that bass contributes less to the measured loudness than it does to a flat RMS reading, and the upper midrange contributes more. This is the main reason a LUFS number and a peak or RMS number tell you different things about the same file. See LUFS vs. dBFS for that distinction.

Channel weighting matters too: in a surround measurement, the surround channels are given extra weight, and in standard practice the LFE channel is excluded from the loudness sum.

Momentary: a 400 ms window

Momentary loudness averages over a sliding window of roughly 400 milliseconds. It moves fast enough to follow individual words, drum hits, and sudden level changes. Use it to see what just happened. It is not a delivery number — momentary values on normal program material swing over a wide range.

Short-term: a 3-second window

Short-term loudness uses a sliding window of about 3 seconds. This is the reading that corresponds best to how loud a passage feels while you listen to it: a chorus, a verse, a line of dialogue over music. Mix engineers watch short-term to compare sections of a piece against each other.

Meters that display a loudness range figure build it from the distribution of short-term values across a program. That range describes how much the loudness varies, not how loud the program is; it is specified separately from BS.1770 and different tools can label it differently.

Integrated: the whole program, gated

Integrated loudness (sometimes shown as program loudness) averages the entire measured span — a track, an episode, a commercial. It is the number meant by "this master is −14 LUFS."

The integrated measurement is gated: quiet stretches are excluded so that silence and room tone do not drag the average down. BS.1770 specifies two gates — an absolute threshold well below any normal program level, and a relative threshold set a fixed number of Loudness Units below the ungated average of the material that passed the first gate. In practice this means a track with long silences measures close to the loudness of its actual content, not to its average including the gaps.

Because the gate needs the whole program before it can place the relative threshold, integrated loudness is only final when the measurement stops. Mid-file, it is an estimate.

The reading that sits next to LUFS

BS.1770 also defines the true-peak meter, measured in dBTP, which estimates the peak of the reconstructed waveform rather than the highest stored sample. Loudness and true peak answer separate questions and are quoted together: broadcast and streaming delivery specifications typically pair an integrated loudness target with a true-peak ceiling. The ATSC recommended practice for digital television, A/85, documents that pairing and publishes a short loudness and true-peak quick reference alongside the full document.

For the peak side of that pair, see true peak vs. sample peak. For what happens to an integrated number after upload, see loudness normalization on streaming services. And if your meter is labeled LKFS rather than LUFS, see LKFS vs. LUFS — same scale, different name.