Methodology

Every measurement on lufs.to is made with the mastering.to analyzer. Each track is measured in full, as a whole file, and nothing is normalized, trimmed, or adjusted first.

How a track is measured

Audio files are decoded to full-scale PCM without resampling or gain changes, then analyzed with the same engine that powers the mastering.to web analyzer. Loudness follows ITU-R BS.1770 and EBU R128: K-weighted, gated, and integrated over the whole track. See What is LUFS? for the background.

The measurements

• Integrated LUFS. The loudness of the whole track, K-weighted and gated.
• True peak (dBTP). The highest inter-sample peak, estimated from the reconstructed waveform. It can read higher than the sample peak.
• Loudness range (LRA). How much the loudness varies over the track, in loudness units, per EBU Tech 3342.
• Short-term max. The loudest 3-second window in the track, in LUFS.
• Dynamic range. Crest-factor based: the gap between the track's peak and its RMS level, in dB. It is not the same as the DR14 measurement, or as loudness range.

For a related figure, the converter also calculates peak-to-loudness ratio (PLR) from true peak and integrated loudness.

Sources and file formats

Tracks are measured from bit-exact CD rips, vinyl transfers, and lossless or store downloads. Each release shows where its files came from and what format was measured. Lossy files (MP3, AAC) are decoded to PCM and labeled as such. Lossy encoding can shift true peak and, slightly, other values compared with the lossless master, so treat those figures as approximate.

Credits and metadata

Release details, performer and remixer credits, and mastering engineer and studio credits come from the release itself and from published sources such as Discogs and Beatport. Credits that can't be verified are left out, not guessed, so a blank engineer or studio means none was verified.

Comparisons and cohorts

The comparisons on track and release pages, such as "louder than 68% of tracks", are made within a cohort: tracks of the same genre and decade, or by the same engineer or studio. A comparison is only shown when a cohort has at least 10 tracks. Smaller groups are too small to state anything about responsibly.

Limitations

Each track is measured once, from one specific release. Other masters, reissues, and streaming versions of the same music can measure differently. The database is hand-curated and doesn't aim to be exhaustive.

See also: Terms for our accuracy disclaimer, and Our Data for how the data can be used.