LUFS, True Peak and Streaming Normalization: What Actually Happens to Your Master

Loudness Is Measured, Not Guessed

Peak meters tell you how close a signal gets to clipping. They say almost nothing about how loud it sounds. LUFS (Loudness Units relative to Full Scale) exist to measure perceived loudness in a standardized way, defined in ITU-R BS.1770 and used by broadcasters and streaming platforms. One LU is one decibel of loudness change, so a master at −9 LUFS is 5 dB louder than one at −14 LUFS.

How the Measurement Works

K-weighting

Before measuring, the signal passes through a filter that approximates how we hear: a high-shelf boost around the upper mids and treble and a high-pass that discounts deep lows. Two mixes with the same RMS level but different tonal balance can measure several LU apart - the brighter, mid-forward one reads louder, because it sounds louder.

Three time windows

• Momentary - 400 ms window. Useful for spotting the loudest moments.
• Short-term - 3 second window. The best read on how loud a section feels.
• Integrated - the whole track. This is the number platforms use.

Gating

Integrated loudness ignores silence and very quiet passages. Blocks below an absolute gate of −70 LUFS are discarded, then a relative gate drops blocks more than 10 LU below the level of what remains. A quiet intro doesn't drag the number down much, and a song with a long silent tail measures the same as without it.

Loudness range

LRA (loudness range) describes how much the short-term loudness varies across the track, in LU. A dense EDM master might sit at 4-5 LU, a dynamic film cue at 15 or more. It's a useful sanity check on how much macro-dynamics your master still has.

Sample Peak vs True Peak

A sample peak meter reports the highest sample value. But the analog waveform that a converter reconstructs between samples can rise above the highest sample - an inter-sample peak. A master whose samples never exceed −0.1 dBFS can produce a reconstructed waveform that goes over 0 dBFS.

True-peak meters estimate this by oversampling the signal (BS.1770 specifies at least 4x at 48 kHz) and measuring the peaks of the oversampled waveform. The result is reported in dBTP.

It matters most after lossy encoding. Converting to AAC, Ogg Vorbis or MP3 removes and reshapes content, and the decoded result usually has higher peaks than the source. A master limited hard to −0.1 dBTP can clip once it's encoded and decoded on a listener's phone. That's why platforms recommend about −1 dBTP, and more headroom for very loud masters.

What Streaming Platforms Do on Playback

Most major platforms normalize: they measure each track's integrated loudness and adjust playback gain toward a reference level. Exact policies change and some depend on the user's settings, so check the platform's current documentation before a release. At the time of writing, the commonly published figures are:

• Spotify - around −14 LUFS by default, with "Loud" and "Quiet" options for listeners. Spotify recommends masters peak no higher than −1 dBTP, or −2 dBTP for masters louder than −14 LUFS.
• Apple Music - Sound Check normalizes to around −16 LUFS.
• YouTube - around −14 LUFS, and it turns loud content down rather than turning quiet content up.

Two details are widely misunderstood:

• Turning down is free, turning up is not. A loud master is simply played quieter. A quiet master is only raised as far as its peaks allow, so a very dynamic master may end up a little quieter than the reference.
• Albums are treated as albums. When a listener plays a full album, platforms with album normalization apply one gain to the whole record, preserving the level differences between your songs. Your quiet interlude stays quiet.

What This Means for Your Master

Loudness no longer buys loudness

If your master is at −8 LUFS and a competitor's is at −13, both will play back at roughly the same level on a normalized platform. What you keep is the cost of getting to −8: reduced transients, more limiting artifacts and a flatter-sounding track next to the one that kept its punch. Played side by side at matched loudness, the more dynamic master often sounds bigger.

But loudness is still a genre choice

Density is part of how some genres sound. Modern club music, metal and some hip-hop are expected to feel compressed and full. Normalization doesn't make that wrong - it makes it a sound decision instead of a competitive one. Master for the character the music needs, then confirm it survives normalization.

A sensible workflow

1. Mix without a limiter on the master bus, or with a gentle one, and leave headroom.
2. Master for the sound you want. Watch short-term loudness in the loudest section rather than chasing an integrated number.
3. Set the final limiter's ceiling to about −1 dBTP with true-peak detection on. Go lower if the master is very loud.
4. Check a lossy encode: export to AAC or MP3 and measure the true peak of the decoded file.
5. Compare against references at matched loudness. Most loudness meters can do this, and it's the only fair comparison.

The number on the meter is a measurement, not a target. Use it to make sure your master translates, not to decide how it should sound.