What exactly is a LUFS?

It's a unit that measures perceived loudness, not electrical amplitude. It accounts for the ear not being equally sensitive to all frequencies and averages over time, so it comes reasonably close to what a person would call «louder». That's why it has replaced peaks as the delivery reference.

What level should I deliver at?

Between -14 and -9 integrated LUFS covers most current cases well. More important than the exact number is that the master sounds good at that level: a track at -12 LUFS with healthy dynamics beats one at -8 that's crushed, because the platform is going to turn the second one down anyway.

What does platform normalisation do?

They measure each track's loudness and adjust playback to a common value, so the listener doesn't have to touch the volume between songs. In practice that means if you deliver louder than their target, they turn you down; and if you deliver quieter, some turn you up and some don't.

So is there no point pushing the master any more?

No point in terms of being louder than everyone else, which was the original driver of the loudness war. What it does still do is change the character: heavy compression adds density and aggression, and that can be a legitimate artistic choice. What no longer makes sense is doing it out of fear of ending up quiet.

What is true peak and why am I asked for -1 dBTP?

The waveform's real peak can exceed the sample values when the signal is reconstructed, and lossy encoders (MP3, AAC) make it worse. Leaving the ceiling at -1 dBTP stops that reconstruction distorting in the listener's player. It's cheap headroom that saves trouble.

Integrated, short-term or momentary LUFS?

Integrated averages the whole song and is what platforms use: it's the one to watch when deciding delivery level. The other two show internal movement, for instance how much the chorus lifts over the verse.

Mastering10 min read

LUFS and loudness: what level to deliver for streaming

What a LUFS measures, what each platform's normalisation does, and why squashing your master no longer makes you louder than anyone.

Why we stopped measuring in peaks

For years level was controlled by looking at peaks, and that had an obvious problem: two songs with the same peak can sound completely different in loudness. A piano recording peaking at 0 dBFS and a compressed rock track with the same peaks are nothing alike to a listener, because what we perceive as volume depends on sustained energy, not on the highest point.

LUFS solves that: it measures perceived loudness, weighting frequencies according to the ear's sensitivity and averaging over time. It's reliable enough that streaming platforms adopted it to level their entire catalogues, and that decision completely changed the arithmetic behind a master.

What happens when you upload

Platforms measure your track's integrated loudness and adjust playback to bring it near their reference value. If you delivered above it, they turn you down; if you delivered below, some turn you up and some leave you alone. Either way the listener hears every song at a similar volume without touching anything.

Platform	Approximate reference	What it means
Spotify	≈ -14 LUFS	Turns down anything louder; with normalisation on, pushing gains you nothing
Apple Music	≈ -16 LUFS	Turns both up and down to match; rewards preserved dynamics
YouTube	≈ -14 LUFS	Turns down anything louder and doesn't compensate upwards
Tidal / Amazon	≈ -14 LUFS	Behaves much like the rest

These figures change from time to time and each platform tunes them to its own criteria, so treat them as an order of magnitude rather than a specification. What doesn't change is the mechanism, and the mechanism is what dictates the strategy.

The loudness war is over, and nobody won

For two decades there was a real reason to push: on radio or on a CD, the louder track stood out. That reason disappeared with normalisation. Today the only consequence of delivering very loud is that the platform turns you down, and the listener gets a flattened version playing at the same volume as everything else — the worst of both worlds.

That doesn't mean compression is wrong. Density, aggression and constant push are perfectly valid aesthetic choices, and whole genres are built on them. The difference is the motive: compressing because you want that character is legitimate; compressing out of fear of sounding quieter than someone else is throwing away dynamics for nothing.

How to decide your delivery level

Check the integrated LUFS of two or three commercial references in the style. That range is your realistic target.
Leave true peak at -1 dBTP. It's cheap headroom and avoids distortion when encoding to MP3 or AAC.
Check the dynamic range: if chorus and verse read almost the same, you've flattened the song's structure.
Always compare at matched loudness. Without that, louder will seem better and you'll decide badly.
If you're torn between two levels, deliver the quieter one: a platform can turn it up, but it can't give dynamics back.

The last point saves the most arguments. And if you're getting to this level of detail, it's worth having the earlier stages settled: the mastering guide covers the full chain, and the gain staging one covers the headroom the mix should arrive with.

Frequently asked questions

What exactly is a LUFS?: It's a unit that measures perceived loudness, not electrical amplitude. It accounts for the ear not being equally sensitive to all frequencies and averages over time, so it comes reasonably close to what a person would call «louder». That's why it has replaced peaks as the delivery reference.
What level should I deliver at?: Between -14 and -9 integrated LUFS covers most current cases well. More important than the exact number is that the master sounds good at that level: a track at -12 LUFS with healthy dynamics beats one at -8 that's crushed, because the platform is going to turn the second one down anyway.
What does platform normalisation do?: They measure each track's loudness and adjust playback to a common value, so the listener doesn't have to touch the volume between songs. In practice that means if you deliver louder than their target, they turn you down; and if you deliver quieter, some turn you up and some don't.
So is there no point pushing the master any more?: No point in terms of being louder than everyone else, which was the original driver of the loudness war. What it does still do is change the character: heavy compression adds density and aggression, and that can be a legitimate artistic choice. What no longer makes sense is doing it out of fear of ending up quiet.
What is true peak and why am I asked for -1 dBTP?: The waveform's real peak can exceed the sample values when the signal is reconstructed, and lossy encoders (MP3, AAC) make it worse. Leaving the ceiling at -1 dBTP stops that reconstruction distorting in the listener's player. It's cheap headroom that saves trouble.
Integrated, short-term or momentary LUFS?: Integrated averages the whole song and is what platforms use: it's the one to watch when deciding delivery level. The other two show internal movement, for instance how much the chorus lifts over the verse.