Free Online LUFS Meter

Integrated LUFS — the gated, perceived loudness of the whole track per ITU-R BS.1770. This is the value platforms use to decide whether to turn your track up or down.

Loudness Range (LRA) — dynamic range of loudness, in loudness units. Low LRA = squashed master, high LRA = living dynamics.

True Peak (dBTP) — the maximum inter-sample peak after oversampling (at least up to an equivalent of 192 kHz; 4× for 48 kHz material). Keep it below the platform ceiling to avoid clipping when transcoded to lossy formats.

Short-term max / Momentary max — the loudest 3-second / 400-ms window. Useful for spotting the hottest sections.

Back in the 80s, studio loudness was measured on VU meters and peak meters: voltage, RMS, dBFS. That was enough while each release sat on a single medium — vinyl, cassette, later CD — and its loudness was baked into the pressing.

With CDs in the 90s came the loudness war. Producers noticed: a track that plays louder than its neighbors on radio sounds brighter and grabs attention. To win, each new release was squeezed harder — compressors, limiters, masters pushed almost to 0 dBFS. Over a decade the average loudness of pop masters went up 6–10 dB, but dynamics died with it: quiet passages became as loud as the loud ones.

Then streaming arrived in the 2010s and solved the war with a single rule: we normalize loudness at playback. If you deliver a track at −6 LUFS, Spotify will pull it down to −14 anyway — all your loudness battle evaporates, leaving only a crushed sound.

That’s when the industry needed an honest, reproducible loudness metric — not dBFS, not RMS, not peaks. The answer was ITU-R BS.1770 and its unit LUFS. It measures how loud a track actually sounds to the ear, not how close it sits to the digital ceiling.

The ITU-R BS.1770 standard first appeared in 2006; the current version is BS.1770-5 (2023). It describes how to measure loudness properly. Three steps.

1. K-weighting — two filters in series: a pre-filter (high-shelf, ~+4 dB above 2 kHz) and an RLB filter (high-pass with noticeable roll-off below 100 Hz). This is a rough approximation of human hearing — the ear is less sensitive to low frequencies and more sensitive to the upper mid. The filtered signal is then mean-squared.

2. Time windows. Momentary is averaged over 400 ms, Short-term over 3 seconds, Integrated over the whole track with gating. Small windows catch momentary perception spikes; the big one gives overall loudness.

3. Gating. When computing Integrated LUFS the standard excludes quiet blocks — first anything below −70 LUFS (absolute gate), then anything below −10 LU relative to the loudness measured after the absolute gate (relative gate). Without gating, silence at the beginning and end would drag the average down by several LU.

True Peak is defined in BS.1770 Annex 2 — a peak value measured after oversampling to at least an equivalent of 192 kHz (4× for 48 kHz; 2× is enough for 96 kHz). That’s enough to catch intersample peaks. EBU Tech 3341 standardizes the meter requirements.

All four are "loudness", but they measure different things. Easy to confuse, especially coming from a peak meter.

In short: dBFS tells you "how close to the digital ceiling", LUFS tells you "how loud to the ear". Streaming cares about LUFS; clipping protection cares about True Peak. RMS is now a historical metric.

If the file on disk peaks at −0.1 dBFS, it’s tempting to assume it doesn’t clip. It does. After D/A reconstruction at playback, an intermediate value between two "honest" samples can exceed 0 dBFS. That’s an intersample peak. The speaker doesn’t know, the amp doesn’t always know either — but a lossy decoder can reconstruct that peak explicitly and clip.

Typical behaviour:

That’s why streaming services set the ceiling at −1 dBTP (or −2 for "loud" presets) rather than 0 — headroom for the codec, so there’s no audible clipping on listeners' devices after transcoding to AAC/OGG. CD and Bandcamp play the uncompressed file as-is — no transcoding there, but even then 0 dBFS sample peak does not guarantee there are no intersample clips after D/A reconstruction in the listener’s player.

Simple mastering rule: if the file goes to streaming — set the limiter ceiling to −1 dBTP. For CD or Bandcamp — at least −0.3 dBTP as insurance against intersample peaks; for very loud masters you can stick to −1 dBTP.

Different destinations treat your file differently. The analyzer splits them into three groups because the rules differ in principle:

Loudness normalization isn’t new. The idea came from ReplayGain (2001), which proposed metadata in audio tags. Apple shipped Sound Check in iTunes in 2002, and in 2014 Spotify launched modern BS.1770-based normalization. Today almost every major service uses the same algorithm, with different targets and logic.

Spotify (official guide): default target −14 LUFS (Normal mode). Users can switch to Loud (target −11 LUFS) or Quiet (target −19 LUFS). If the track is above target, the platform reduces gain at playback; if it’s quieter and Loud is selected, positive gain is added with an internal limiter around −1 dBTP to pull a quiet track up to target.

Apple Music — Sound Check, target −16 LUFS. The algorithm is very close to BS.1770, though Apple historically doesn’t publish the details. In practice the numbers match BS.1770 within 0.1–0.2 LU.

YouTube / YT Music: normalization down only, to −14 LUFS. A quiet track will stay quiet relative to its neighbors — YouTube won’t pull it up. This matters for albums: if one track is louder than the others, it gets lowered, and the quiet ones don’t rise to match.

Tidal, Amazon Music: −14 LUFS, logic similar to Spotify.

Deezer: −15 LUFS.

SoundCloud: nominally recommends −14 LUFS but actually plays files without normalization. That makes peaks here as important as loudness.

Important: the numbers below are aesthetic master levels BEFORE normalization, not platform requirements. Platform demand is uniform: −14 LUFS (for most). Genre values exist so the track sounds dense and alive after streaming pulls it to its target.

After normalization the integrated loudness will still be matched to the platform target, but the living dynamics (LRA) and character stay — whereas a master crushed to −8 LUFS will, after Spotify, lose everything you fought for.

Tracks with low LRA sound more aggressive and dense after normalization, but without dynamics. High-LRA tracks breathe, keep the contrast between loud and quiet parts. If your genre is classical or jazz, crushing to −8 LUFS is pointless: after Spotify the track will drop ~15 LU and lose everything you fought for.

If you approach the final normalization with the right level structure, the master limiter works almost invisibly. A few simple rules:

The file never leaves the browser. Decoding and measurement run locally via the Web Audio API. No analytics are loaded inside the analyzer.

Use the mastering tool to normalize your track to any of these targets — loudness, true-peak ceiling and compression in one pass.