What Is LUFS?

LUFS shows up on every streaming platform's loudness guidance and every modern track analyzer, usually with no explanation beyond "aim for around -14." Here's what the number actually measures and why it exists.

The standard behind it

LUFS (Loudness Units, Full Scale) is the unit used by ITU-R BS.1770, the standard that defines how to measure a track's perceived loudness algorithmically. It's sometimes written LKFS (Loudness, K-weighted, Full Scale) — same measurement, different name, mostly a US-broadcast-vs-international naming split rather than a technical difference.

The point of BS.1770 was solving a real problem: a plain peak or RMS meter doesn't agree with how loud something actually sounds to a human ear, because human hearing isn't equally sensitive across all frequencies, and a single very loud instant doesn't make a whole song feel loud. BS.1770 does two specific things differently:

1. K-weighting. The signal is run through a filter that de-emphasizes very low frequencies before measurement — modeling, roughly, how the ear (and head/torso, acoustically) responds across the spectrum. A sub-heavy track doesn't get an inflated loudness reading just because there's a lot of low-frequency energy the filter isn't as sensitive to.
2. Gating. Loudness is measured in 400ms blocks, and quiet blocks are excluded from the final average — first anything below an absolute floor of -70 LUFS, then anything more than 10 LU quieter than the (already partially gated) running average. This stops a song with long quiet intros or breakdowns from measuring artificially loud just because the loud sections are loud — a plain average without gating would be dragged down by the silence and misrepresent the song's actual perceived level.

The actual formula

Per-block loudness is:

L = -0.691 + 10 × log₁₀(mean square of the K-weighted signal)

Integrated (whole-track) LUFS is the loudness of the average mean-square power across every block that survives both gates — not a simple average of the per-block LUFS numbers themselves (averaging in the log domain vs. the power domain gives different, wrong, results, which is a common implementation bug in home-rolled loudness meters).

A useful sanity check if you ever want to verify a meter is behaving correctly: a full-scale (0 dBFS), 1 kHz sine wave measures at very close to -3.7 LUFS integrated. That number is specific enough, and consistent enough across correct implementations, that it's a genuinely useful way to spot-check a tool — including this site's own analyzer.

Integrated vs. short-term vs. momentary

Analyzers usually report three variants:

• Integrated LUFS — the whole-file average described above. What streaming platforms normalize against.
• Short-term LUFS — a rolling 3-second window. Useful for seeing how loudness moves through a track section by section.
• Momentary LUFS — a rolling 400ms window, essentially the raw per-block value before any gating or averaging. The most reactive, least representative of the whole track on its own.

Why it matters practically

Spotify, Apple Music, YouTube, and Tidal all normalize playback loudness — turning louder tracks down (and, on most platforms, quieter tracks up) toward a target integrated LUFS value so a whole playlist doesn't swing wildly in volume between tracks. The specific targets each platform uses do shift over time as platforms retune their normalization, so treat any single quoted number as a snapshot, not a permanent spec — but the practical consequence stays the same either way: mastering drastically louder than a platform's target doesn't make your track sound louder to a listener on that platform, it just gets turned down further, while the extra limiting you used to get there has already cost you dynamic range you can't get back.

Check your own track

Analyze My Track runs a real BS.1770-4 measurement — integrated, short-term, and momentary LUFS together with true peak and dynamic range — on your actual upload, free, so you're working from a real number instead of guessing.