Loudness & LUFS

How loud should your master be? For years the answer was "as loud as possible". Streaming changed that. Learn how loudness is measured, what streaming platforms do with it, and how to pick a level that sounds great everywhere.

The Loudness War

When two songs play one after another, the louder one usually sounds better at first: fuller, brighter, more exciting. Producers and labels noticed this. From the 1990s through the 2000s, masters got louder and louder to stand out on radio, CD changers and playlists. This race is called the loudness war.

Because digital audio can't go above 0 dBFS, the only way to make a song louder is to reduce its dynamics (the difference between quiet and loud moments) with heavy compression and limiting. Extreme examples ended up with flattened drums, audible distortion and listening fatigue. Some heavily criticized albums from that era, such as Metallica's Death Magnetic (2008), became famous examples of over-limiting.

Today, most streaming platforms use loudness normalization, which removes most of the benefit of an extremely loud master. More on that below.

Peak, RMS & LUFS

There are different ways to measure level. Each tells you something different.

Measurement	What it measures	Good for
Peak (dBFS)	The single highest sample level	Avoiding clipping. Says almost nothing about how loud it sounds.
RMS (dB)	The average level over a short window	A rough idea of perceived loudness. Treats all frequencies the same.
LUFS	Loudness Units relative to Full Scale. An average weighted to match human hearing	The industry standard for perceived loudness and streaming targets

A snare hit and a sustained pad can have the same peak level but very different loudness. That's why peak meters can't tell you how loud a song sounds.

LUFS is defined by the international standard ITU-R BS.1770 (also used in EBU R128). It applies a filter called K-weighting that reduces deep bass and slightly boosts the upper mids, roughly matching how our ears respond. 1 LU equals 1 dB, so a track at -10 LUFS is 4 dB louder than one at -14 LUFS.

The three LUFS time windows

Momentary, short-term, integrated

Momentary 400 ms window jumps around with every hit Short-term 3 s window shows the loudness of a section (verse vs chorus) Integrated whole song one number for the entire track, ignoring silence

Integrated LUFS is the number streaming platforms use. It uses a "gate" so silence and very quiet parts don't drag the average down. Short-term LUFS is useful while mastering: watch the loudest chorus to see how hard you're pushing.

Note

The demo on this page shows peak and RMS meters, not LUFS. RMS behaves in a similar way for this experiment: watch the gap between peak and RMS shrink as you push harder. For real masters, use a proper LUFS meter.

True Peak & Loudness Range

True peak (dBTP)

Digital audio stores samples, but the analog wave that comes out of a speaker is smooth. Between two samples, the real waveform can rise above the highest sample. These are called inter-sample peaks.

A true-peak meter estimates those hidden peaks and shows them in dBTP (decibels true peak). A file can read 0.0 dBFS on a sample peak meter but +1.0 dBTP on a true-peak meter. When that file is converted to analog or encoded to MP3/AAC for streaming, the hidden peaks can clip.

That's why streaming masters typically use a limiter ceiling of -1 dBTP, and very loud masters often use -2 dBTP to stay safe.

Loudness range (LRA)

Loudness range (LRA), measured in LU, shows how much the loudness varies through a song. It ignores the loudest and quietest extremes, so a single drum hit or fade won't change it much.

Low LRA (only a few LU): very consistent loudness. Typical of loud EDM, pop and hip-hop.
High LRA: big differences between quiet and loud sections. Typical of film scores, classical and acoustic music.

Neither is "right". A ballad that builds from a quiet piano to a full band should have more loudness range than a club track.

LUFS stands for Loudness Units relative to .

Streaming platforms use the LUFS value of a whole song.

Peaks between samples are measured in .

Streaming Normalization

Loudness normalization means the platform measures each song's loudness and adjusts its playback volume, so songs play at a similar level. If your master is louder than the platform's target, it gets turned down.

Platform	Approximate reference level
Spotify	about `-14 LUFS` integrated (default setting)
Apple Music	about `-16 LUFS` (with Sound Check on)
YouTube	about `-14 LUFS`

These numbers change

Treat these as approximate targets, not rules. Platforms adjust their systems, offer user settings (Spotify, for example, has louder and quieter volume-level options), and handle quiet tracks differently: some turn them up, some don't. Check each platform's current documentation before you release.

What normalization means for you

Imagine two masters of the same song:

Two masters on a -14 LUFS platform

Master A -8 LUFS heavily limited → turned down 6 dB → plays at -14 Master B -13 LUFS dynamic, punchy → turned down 1 dB → plays at -14

Both now play at the same loudness. But Master A has flattened transients and less punch, while Master B keeps its impact. Next to each other, Master B often sounds bigger and clearer, even though it was the "quieter" master.

This is why loud masters can sound worse after normalization. The loudness advantage is gone, and only the side effects of heavy limiting are left.

Compare Fairly: Level-Matching

The single most important habit in mastering is level-matching: comparing two versions at the same loudness. Without it, you'll always prefer the louder one, and you'll keep pushing the limiter harder.

What to listen for in the demo:

With Compare fairly: No, more drive sounds "better" for a while, purely because it's louder.
With Compare fairly: Yes, the output level stays roughly the same. Now you hear only the cost of limiting: softer drum hits, less space and, at high drive, distortion.
Watch the meters. The peak stays pinned near the ceiling while the RMS rises toward it. The smaller the gap between them, the less dynamic the track.

Producer tip

Many limiters and mastering suites have a level-match or "gain match" button that automatically lowers the output while you compare. Use it every time you judge a change.

Dynamics & Choosing a Target

So how loud should you master? There's no single right number. Consider:

Genre: modern EDM, trap and metal are often mastered louder than -14 LUFS because the style relies on a dense, intense sound. Jazz, folk and classical are usually more dynamic.
References: measure the short-term loudness of the chorus of a professional reference in your genre. That tells you what the style expects.
The song itself: push until it starts to sound worse at matched level, then back off a little.

A useful related measure is PLR (peak-to-loudness ratio): the true peak minus the integrated loudness. A master peaking at -1 dBTP with -14 LUFS has a PLR of 13 dB. A master at -1 dBTP and -7 LUFS has only 6 dB, which means much less room for transients.

Metering Tools

Stock meters: Logic Pro includes a Loudness Meter plugin. Check your DAW's manual for its loudness metering options.
Free meters: Youlean Loudness Meter has a widely used free version with integrated, short-term, momentary, true peak and LRA readings.
Mastering tools: FabFilter Pro-L 2 and iZotope Ozone (and its metering tools) show LUFS and true peak alongside the limiter.

Place the loudness meter last in the chain, after the limiter, so it reads the final output.

Try it in your DAW

Put a loudness meter on the master track [Ableton: Main; FL Studio: Master; Logic: Stereo Out], after any limiter.
Play your whole song from start to finish and write down the integrated LUFS and the maximum true peak.
Import a professional reference in your genre. Measure its integrated LUFS the same way.
Make two versions of your master: one at about -14 LUFS and one about 4 dB louder. Keep the ceiling at -1 dBTP.
Turn the louder version down by the difference so both measure the same. Listen and decide which has more punch.

1. Why can't a peak meter tell you how loud a song sounds?

Peak meters only work in mono
It shows only the single highest sample, not the average energy our ears hear
Peak meters measure frequency, not level
Peak meters are always inaccurate

Two songs with the same peak can sound very different in loudness. LUFS measures perceived loudness over time.

2. A master at -8 LUFS is played on a platform with a -14 LUFS reference. What happens?

It plays 6 dB louder than other songs
It is rejected
It is turned down by about 6 dB
It is converted to mono

Normalization turns loud tracks down to the reference, so the extra loudness is lost while the side effects of limiting remain.

3. What is a true-peak meter designed to catch?

Inter-sample peaks that can clip after conversion or encoding
Phase problems between left and right
Sibilance in vocals
Low-frequency rumble

The real waveform can go above the highest sample. True-peak meters estimate these peaks in dBTP.

4. Which LUFS window is about 3 seconds long?

Momentary
Integrated
True peak
Short-term

Momentary is 400 ms, short-term is 3 s and integrated covers the whole song.

5. Why should you level-match before comparing two masters?

To save CPU
Because streaming services require it
To raise the true peak
Louder sounds better at first, so a fair comparison needs equal loudness

Without level-matching, you will keep choosing the louder version, even when it is actually worse.