LUFS explained: how loudness is measured and which targets to use

LUFS (loudness units relative to full scale) is a measure of how loud audio sounds to people, rather than how high its peaks are. Broadcasters and streaming services use it to play everything at a similar level: European broadcast aims for −23 LUFS, and most music streaming services turn tracks to somewhere around −14 to −16 LUFS. The loudness meter measures a file the same way, in your browser, and compares it with those targets.

What LUFS measures

The method is defined in ITU-R BS.1770 and adopted by the European Broadcasting Union in EBU R128. Two filters first apply K-weighting: a gentle high-pass that ignores deep rumble and a high-shelf boost of about 4 dB above roughly 1.5 kHz, approximating how the head and ear respond. The meter then takes the average power of the filtered signal in 400 millisecond blocks, overlapping by 75 percent, and expresses it in decibels. One loudness unit (LU) equals one decibel, so a track at −10 LUFS is 4 dB louder than one at −14 LUFS. LUFS and LKFS, the American name, are the same unit. Because the measurement follows perception, a bass-heavy track and a bright one with the same reading should sound about equally loud, even if their waveforms look very different.

Integrated, short-term and range

A meter reports several numbers:

• Integrated loudness is the whole programme. Blocks quieter than −70 LUFS are ignored, then any block more than 10 LU below the average of the rest is ignored too. This gating stops silence and quiet passages from dragging the figure down.
• Momentary loudness uses a 0.4 second window and short-term loudness a 3 second window; the meter here reports the maximum of each.
• Loudness range (LRA) describes how much the level varies, from the 10th to the 95th percentile of short-term values. A compressed pop master may show 4 to 6 LU; a film or classical recording can exceed 15.

True peak and why it matters

Sample peak is the highest sample value in the file. The real waveform, once converted back to analogue or decoded from MP3 or AAC, can rise slightly higher between samples. True peak, in dBTP, estimates that by oversampling. EBU R128 sets a maximum of −1 dBTP for production, and many streaming services suggest −1 dBTP or lower, because lossy encoding adds small overshoots. The meter on this site oversamples four times, which is the standard approach, and I describe its true peak as an estimate within a few tenths of a decibel.

Loudness targets

• EBU R128 broadcast: −23 LUFS integrated, −1 dBTP maximum.
• US television (ATSC A/85): −24 LKFS.
• Spotify, YouTube and Amazon Music: around −14 LUFS.
• Apple Music: around −16 LUFS.
• Podcasts: commonly −16 LUFS for stereo and −19 for mono.

Treat the streaming figures as a snapshot. Platforms change their reference levels and options, some only turn loud tracks down rather than turning quiet ones up, and listeners can switch normalisation off. With normalisation on, mastering much louder than −14 LUFS no longer makes a song louder on most services; it just gets turned down and loses dynamics. The history of that change is told in Audio normalization explained: peak, RMS and loudness, and why streaming turns you down.

How to hit a target

Measure first, then apply gain equal to the target minus the measurement. A podcast at −20.5 LUFS needs 4.5 dB of gain to reach −16. Before raising it, add the gain to the true peak: if the result goes above −1 dBTP, a limiter must catch the peaks, otherwise the file will clip. The meter flags this in its table. Turning a file down never risks clipping. Note that the audio normalizer on this site works on peaks, not loudness: it brings the highest peak to a chosen level, which is useful before encoding but does not set LUFS by itself. When you export, keep the true peak a little lower for MP3, since How MP3 compression works, and which bitrate you can actually hear explains why encoding can add overshoots, and save with the MP3 converter. Measure the final file again; it takes seconds and confirms the result.