Guide · Analysis
LUFS Explained: Loudness, Targets and Normalisation
Decibels measure level. LUFS measures loudness. The difference is not academic: it is the difference between a master that is technically correct and one that plays at the right volume on the platform your listeners are using.
Why Peak Level Is Not Loudness
The oldest measurement in audio is peak level: the highest single sample in the file, expressed in decibels below full scale. It is precise, it is easy, and it tells you almost nothing about how loud a track will sound.
This is because the ear does not weigh all frequencies equally. A 1 kHz tone at −20 dBFS is heard as louder than a 50 Hz tone at −20 dBFS, by a large margin. A track with a lot of bass energy can have a modest peak level and still sound extremely loud, because a small part of the signal is carrying a disproportionate amount of the perceived loudness.
LUFS, short for Loudness Units relative to Full Scale, corrects for this. The unit is deliberately the same as decibels, so the numbers are familiar, but the measurement is perceptual rather than electrical.
How LUFS Is Measured
The measurement is defined by ITU-R BS.1770, and it is a four-step process.
- K-weighting. The signal goes through a two-stage filter: a high-shelf that boosts around 1.7 kHz, modelling the head's effect on sound arriving through the skull, and a high-pass at about 38 Hz that removes the rumble below the standard's cut-off. The result approximates the frequency response of human hearing.
- Block measurement. The filtered signal is measured in overlapping 400 millisecond blocks, stepping forward 100 milliseconds at a time.
- Gating. This is the part most people miss. Blocks below −70 LUFS are discarded as silence. Then the loudness of what remains is calculated, and a second gate removes anything more than 10 dB below that. The result is a measurement that ignores silence and quiet passages, which is what a listener does.
- Summation. The gated blocks are averaged and converted to LUFS.
The gating is why a track with a long quiet intro and a loud chorus measures the same as the same track without the intro. Without it, the intro would drag the number down and the track would play quietly on the platform.
The Numbers You Will See
| Term | What it means |
|---|---|
| Integrated LUFS | The loudness of the whole track, gated. This is the number that matters for delivery. |
| Momentary | The loudness over a 400 millisecond window, as it would be measured live. |
| Short-term | The loudness over a 3 second window, which catches sections rather than hits. |
| Loudness range (LRA) | The spread between the quiet and loud parts. Under 6 LU is very dynamic, over 12 LU is very compressed. |
| True peak | Not loudness at all: the highest peak after inter-sample reconstruction, which is what a DAC actually outputs. Relevant for preventing clipping, not for loudness. |
The analyser on this site reports integrated LUFS and the loudness range, measured to the same recommendation, so you can check a master or a rough mix before you export it.
Why Streaming Normalises
Platforms play music through the phone in a café, through earbuds on a train, through a Bluetooth speaker in a kitchen and through a home cinema at midnight. A single fixed volume cannot be right for all of them.
So the platforms normalise. They measure the loudness of your track and adjust it, usually towards a target, before playback. The practical consequences are:
- A quiet master will be turned up. If you deliver at −20 LUFS to a platform targeting −14, the listener hears a version 6 dB louder than you made.
- A very loud master will be turned down. Deliver at −8 LUFS to a −14 target and the listener hears it quieter, with whatever artefacts the platform's limiter introduces.
- Relative differences within your catalogue survive. Normalisation moves the whole track, so a quiet song next to a loud one still sounds quiet.
What To Target
The honest answer is that the target depends on the platform and the genre, and that being aware of the number is worth more than the number itself.
- −14 LUFS integrated is the common streaming target, and the reference most services normalise towards.
- −9 to −11 LUFS suits genres where the music is expected to be loud and the platform normalises less aggressively.
- −16 to −18 LUFS is safer for material that will also be released on CD, vinyl or in a film, where there is no normalisation at all and the master is played as authored.
What you should not do is aim for the loudest possible master and rely on the platform to turn it down. Every decibel you push past the target is a decibel of limiting the platform will apply for you, and its limiter is not as good as yours.
LUFS Is Not a Target You Set, It Is a Measurement You Read
This is the most useful thing to understand. LUFS does not change how your track sounds. It tells you how loud it is, in a way that accounts for what the ear is doing.
Which means it is a diagnostic as much as a delivery number. If a mix measures −22 LUFS while the material sounds reasonably balanced, you have a mix with a lot of headroom, and you should raise the level rather than compress it further. If a track measures −14 but sounds thin and lifeless, the loudness is being carried by a narrow band and the rest of the spectrum has been squeezed away.
Measure after each stage of mixing and the number becomes a record of your decisions. A mix that was −18 LUFS before the bus compressor and −16 after tells you the bus compressor did 2 dB of work, which is more informative than a number alone.
Common Mistakes
Confusing LUFS with peak level. A track at −1 dBFS peak and −20 LUFS is quiet. A track at −3 dBFS peak and −9 LUFS is loud. The peak tells you how close to clipping you are; the LUFS tells you how it will sound.
Normalising your master to the platform's target. The platform does this itself. If you pre-normalise, you are applying the same attenuation twice.
Chasing a number with compression. If the mix is 6 dB too quiet, adding compression to get there will cost you dynamics. Raise the output instead.
Ignoring the loudness range. A track with a loudness range of 2 LU is essentially compressed flat, whatever its integrated number. The range tells you about the internal dynamics; the integrated number does not.
Questions
What is LUFS?
LUFS is a measure of perceived loudness, in units matching the decibel. Unlike peak level, it accounts for the way human hearing weights different frequencies, so a 1 kHz tone and a 50 Hz tone of the same peak level do not measure the same.
What loudness should I master to?
For streaming, −14 LUFS integrated is the common target, and −9 to −11 for louder genres where the platform does not normalise as hard. The important thing is not the number itself but knowing whether your target is normalised on playback.
Does LUFS change if I change the EQ?
Yes, which is the point of the measure. Cutting 3 dB at 2 kHz changes the perceived loudness even though the peak level is unchanged, because the frequency weighting moves the measurement.
Is LUFS the same as RMS?
No. RMS is a raw average of the signal. LUFS applies a K-weighting filter that models how the ear perceives loudness, then gates the result. A track with a lot of low end will measure considerably louder in LUFS than in RMS.