Two tracks can peak at exactly the same level and still sound nothing alike in volume. One is a sparse piano piece that touches 0 dBFS on a single hammer strike. The other is a dense, limited rock master that sits near the ceiling the whole way through. A peak meter says they match. Your ears say otherwise.
Loudness meters exist to close that gap. Instead of reporting the highest sample, they estimate how loud a listener actually perceives the audio to be. The unit they use is LUFS, and it now underpins broadcast delivery, streaming normalization, and a lot of everyday mixing decisions.
What is LUFS?
LUFS stands for Loudness Units relative to Full Scale. It comes from the ITU-R BS.1770 standard, which describes how to measure program loudness, and the European Broadcasting Union’s EBU R128 recommendation, which builds a practical delivery standard on top of it.
You will also see LKFS (Loudness, K-weighted, relative to Full Scale). It is the same measurement under a different name. North American broadcast specs tend to say LKFS, while music and European broadcast usually say LUFS.
A related unit is the LU, or loudness unit. LUFS is an absolute level, referenced to digital full scale. LU is a relative difference between two loudness readings, and the scale is built so that turning a signal up or down by 1 dB moves its reading by exactly 1 LU. If a song measures -11 LUFS and a service normalizes to -14 LUFS, it will be turned down by 3 dB.
LUFS is not the same as dBFS, though. Because of K-weighting, a track’s LUFS reading depends on its frequency content as well as its level, and two tracks 3 LU apart will not necessarily be 3 dB apart on a peak or RMS meter.
How a loudness meter differs from a peak meter
A peak meter shows the highest instantaneous level. That is essential for avoiding clipping, but it tells you very little about how loud something feels. A single transient can hit the top of the scale while the rest of the track sits far below it.
An RMS meter gets closer by averaging the signal’s power over time, but it treats all frequencies equally. Human hearing does not. We are more sensitive to the upper midrange than to deep bass, so a bass-heavy mix and a bright mix can show the same RMS and sound quite different.
A LUFS meter handles this with K-weighting: a filter applied before measurement. It gently boosts the upper frequencies, roughly modeling the acoustic effect of the head, and rolls off the low end, where our ears are less sensitive. The weighted signal is then averaged. The result tracks perceived loudness far better than peak or plain RMS readings.
Show data
| Frequency | Gain |
|---|---|
| 20 Hz | −13.3 dB |
| 50 Hz | −3.9 dB |
| 100 Hz | −1.1 dB |
| 200 Hz | −0.3 dB |
| 500 Hz | +0.0 dB |
| 1 kHz | +0.7 dB |
| 2 kHz | +3.1 dB |
| 5 kHz | +4.0 dB |
| 10 kHz | +4.0 dB |
| 20 kHz | +4.0 dB |
Momentary, short-term, and integrated loudness
Most loudness meters, including How Loud, show several readings at once.
Momentary loudness averages over a 400 ms window. It moves quickly and reacts to individual hits and phrases. It is useful for spotting a vocal line or a drop that jumps out, but too twitchy to mix to.
Short-term loudness averages over three seconds. It follows the shape of a song section by section, so you can see the verse sitting lower than the chorus or a bridge that loses energy. This is usually the most helpful reading while mixing.
Integrated loudness measures the whole track from the moment you start the meter. It is the single number that delivery specs and streaming services care about. To keep long silences and quiet passages from dragging the result down, integrated loudness is gated: anything below -70 LUFS is ignored outright, and a second relative gate excludes material more than 10 LU below the ungated average. Reset the meter, play the whole song from start to finish, and read the integrated value at the end. Some DAWs, such as REAPER, can very quickly analyze your track and provide a detailed overview of loudness, and potential problems.
Loudness range and peaks
Loudness range (LRA), measured in LU, describes how much the loudness varies across a piece. It is calculated from the distribution of short-term readings, ignoring the very quietest and loudest moments, so a single explosion or a long fade does not skew it. A heavily compressed pop master might have an LRA of a few LU. A film score or a classical recording can easily exceed 15.
LRA is not a quality score. A low number is not wrong and a high number is not better. It is a description of dynamics, and it is useful for checking whether a master matches the character you intended, or whether it will survive a listening environment like a car or a noisy commute.
Loudness normalization changes the level of a whole track, but it does nothing about clipping. Many delivery specs combine a loudness target with a peak ceiling, often expressed in dBTP (decibels true peak). A true-peak measurement estimates the level between samples, where the reconstructed analog waveform can overshoot the highest sample value. That is why a master that never exceeds 0 dBFS on a sample-peak meter can still clip once converted or encoded to a lossy format.
Loudness targets for streaming and broadcast
Most streaming services can normalize playback, measuring each track and adjusting its volume toward a reference level. Broadcasters go further and require programs to hit a specific target. Commonly cited reference levels include:
| Destination | Integrated loudness |
|---|---|
| EBU R128 broadcast (Europe) | -23 LUFS |
| ATSC A/85 broadcast (US) | -24 LKFS |
| Spotify (normalization on, Normal setting) | -14 LUFS |
| YouTube | around -14 LUFS |
| Apple Music | -16 LUFS |
Platforms change their settings from time to time, and some let listeners choose a different normalization level, so check the current documentation before delivering anything with a hard spec. Spotify’s loudness normalization guide is a good example of what to look for.
For broadcast, the target is a delivery requirement. For music streaming, it is a playback setting, and that difference matters.
Does Spotify want -14 LUFS?
Not really. You will often read that “Spotify wants -14 LUFS”, usually followed by the advice to master to -14 and call it a day. But -14 LUFS is the level Spotify’s volume normalization turns tracks toward when a listener has it enabled on the Normal setting. It is not a delivery spec, and nothing is rejected for being louder.
A track mastered at -9 LUFS is simply turned down by about 5 LU. A track mastered at -14 LUFS is left where it is. On paper they now match. In practice, they often do not sound like it.
The reason is that integrated loudness is an average across the whole song. A dynamic master that averages -14 LUFS has quiet verses that sit well below that number and choruses that sit above it. A dense master that averages -9 and gets turned down to -14 keeps its compact shape, so its verses are still close to its choruses. Played back to back, the louder master’s quiet sections can sit noticeably above the dynamic master’s quiet sections, and the denser track can feel bigger even though both measure the same.
Show data
| Section | Dynamic master | Dense master |
|---|---|---|
| Intro | −24.7 | −17.5 |
| Verse | −19.4 | −15.0 |
| Chorus | −11.8 | −12.8 |
| Verse | −19.1 | −15.0 |
| Chorus | −11.7 | −12.8 |
| Bridge | −16.4 | −13.9 |
| Chorus | −11.8 | −12.9 |
| Outro | −24.2 | −19.1 |
Normalization also is not universal. Listeners can turn it off, some choose the Loud or Quiet settings, and other services and players use different references or none at all. When normalization is off, a -14 LUFS master will play back several LU quieter than most commercial releases in the same genre.
So mastering to -14 is not wrong, but it is not a safe default either. For reference, chart pop is mastered far louder. iZotope measured 54 songs that reached the Billboard Global 200 top 10 in the first half of 2024 and found an average of -8.3 LUFS integrated, with most between roughly -9 and -7 and the loudest at -6.
You do not have to follow them there. Our own preference is usually somewhere around -12 to -10 LUFS: dense enough to hold up next to other releases, with enough dynamics left to survive being turned down. That is a matter of taste, not a rule. Some genres push louder still, while acoustic, jazz, classical, and ambient music often sits much quieter, and should.
The right number depends on the song, the genre, and how it was mixed and mastered. A useful approach is to level-match a few commercial references in the same style, measure where they sit, and decide whether your track belongs near them. A loudness meter will not pick the number for you, but it will tell you where you are and how much headroom you may have for normalization.
Why some tracks feel louder at the same LUFS
A LUFS reading is a good estimate of perceived loudness, but it’s only one factor. Two masters at the same integrated level can feel very different, and most of that difference comes from decisions made long before the limiter.
Dynamics and compression. Compression and limiting reduce the gap between the loudest and quietest moments, so more of the track sits near its peak level. That density of sustrained energy reads as loudness. Watch short-term loudness and loudness range together. A track with a low LRA and steady short-term readings will feel loud even after normalization, while a more dynamic track saves its impact for the moments that need it. Neither is automatically better, but push density too far and transients flatten, drums lose their snap, and the track can end up feeling smaller despite measuring louder.
Transients and punch. The opposite also works. A kick or snare with a clear attack can feel louder than a squashed one at the same level, because the ear reads that sharp edge as impact. Transient shaping, careful compressor attack times, and leaving the limiter a little less to do can make a mix feel more energetic without changing the integrated number much.
EQ and the ear’s sensitivity. Our hearing is most sensitive in the upper midrange, roughly 2 to 5 kHz, and least sensitive to deep bass. K-weighting reflects some of that, but EQ still changes how loud a mix feels. Clarity in the presence region makes a track feel forward and loud. Too much of it makes it harsh, and listeners turn it down. At the other end, heavy sub-bass uses up a lot of peak headroom while contributing relatively little to perceived loudness, so the limiter works harder for less result. Tightening the low end, often with a high-pass filter on parts that do not need it, can let the whole mix come up.
Saturation and harmonics. Gentle saturation adds harmonics above the fundamental, where the ear is more sensitive. A bass that is barely audible on small speakers can become clear once it has some upper harmonic content, and a vocal can sit forward without a big level boost. It also rounds off peaks, which gives the limiter a little more room.
A clear mix. When parts fight for the same frequencies, they mask each other, and the mix sounds muddy and smaller. Carving space with EQ and arrangement choices lets each part come through at a lower level, so the track sounds bigger overall.
Contrast. Loudness is relative. A chorus feels loud partly because the verse before it was not. Short-term loudness is the reading to watch here: a few LU of lift into a chorus, through arrangement, automation, or dynamics, often does more for perceived energy than pushing the whole master harder.
Getting a track to feel loud is mostly a mixing job. The meter tells you where you ended up, and helps you check whether a change made the track feel bigger or just measure higher.
How to use a loudness meter while mixing and mastering
Put the meter last in the chain, after any limiter on your master bus, so it reads what you will actually export. Then try a few habits.
Measure the whole song. Reset the integrated reading, play the full track from the start, and note the result. Measuring one loud chorus will overstate the track’s integrated loudness by several LU.
Watch short-term loudness between sections. If the chorus measures only a fraction of an LU above the verse, the arrangement may not be lifting the way you think it is. If the intro sits 10 LU below everything else, a listener with normalization on may reach for the volume knob.
Level-match your references. Comparing your mix with a commercial master is far more honest when both play at the same loudness. Louder almost always sounds better at first, so an unmatched comparison mainly tells you which track is louder. Measure both, then use a gain control to bring them to the same integrated level. How Loud has an onboard gain stage for exactly this kind of quick trim.
Compare before and after processing. When adding a compressor, saturator, or EQ boost, match the output loudness to the bypassed version. It is a fast way to find out whether a plugin is improving the sound or just making it louder.
Leave headroom for encoding. If you deliver near 0 dBFS, lossy encoding can push inter-sample peaks over the top. A ceiling of around -1 dBTP is a common, cautious choice for streaming masters.
Picking a loudness meter
You do not need an elaborate meter to work this way. The essentials are momentary, short-term, and integrated loudness that follow the BS.1770 measurement, a loudness range reading, peak information, and a clear display you can glance at without breaking concentration.
How Loud is our take on that brief: a free, lightweight LUFS meter for macOS, Windows, and Linux in VST3 and AU formats. It shows momentary, short-term, and integrated loudness, levels, peaks, and loudness range, with a gain control for trimming or level matching. It will not tell you what number to master to, but it will show you exactly where you are.
Drop it on your master bus, play your next mix through from start to finish, and see where it lands. Then decide if it sounds good yet.