Every streaming platform turns your music up or down before anyone hears it. That single fact has quietly made the loudness war pointless, and yet producers still crush their masters into a brick because louder used to mean better. It does not any more. If your track measures louder than the platform's target, the platform simply turns it back down, and all you are left with is the damage the limiter did on the way.
This guide covers what LUFS actually measures, what Spotify and Apple Music genuinely do to your file, the true peak number that causes real distortion, and how loud your master should be when you bounce it. Where a platform publishes its own specification we use that; where it does not, we say so, because a lot of the LUFS numbers floating around online are community measurements presented as official policy.
You need a meter that reads integrated LUFS, not your DAW's peak meter. Most DAWs now ship one (Logic's Loudness Meter, Pro Tools' Loudness, Ableton's utility metering), and free plugins like Youlean Loudness Meter cover the rest. A peak meter cannot tell you how loud something sounds.
What LUFS Actually Measures
LUFS stands for Loudness Units relative to Full Scale. It is defined by the ITU-R BS.1770 standard, and unlike a peak meter it tries to measure how loud something sounds rather than how big the numbers in the file are.
It does that with two tricks. First, it applies K-weighting, a filter curve that de-emphasizes low frequencies and slightly boosts the highs, roughly matching how sensitive human hearing is across the spectrum. A 40 Hz sine and a 1 kHz sine at the same peak level do not sound equally loud, and LUFS knows that. Second, it averages over time rather than reacting to instantaneous peaks.
You will see three time windows on a loudness meter:
- Momentary is a 400 millisecond window. It jumps around constantly and is useful for spotting a single loud hit.
- Short-term is a 3 second window. This is the one to watch while a chorus plays, because it tracks the loudness of a section.
- Integrated is the whole track, gated so that silence and very quiet passages do not drag the number down. This is the number platforms use.
One more piece of vocabulary: a Loudness Unit is the same size as a decibel. If your track reads -11 LUFS and the target is -14 LUFS, you are 3 LU too loud, and the platform will apply exactly 3 dB of gain reduction. There is no mystery conversion. Anything you already know about decibels applies here, and our dB Calculator handles the arithmetic if you would rather not do it in your head.
You will also see LRA, or loudness range, on most meters. That is the difference between the quiet and loud parts of your track measured in LU. A heavily limited EDM master might read 3 LU; an orchestral recording might read 15 LU or more. It is a useful sanity check on how much dynamic life is left in a mix.
How Streaming Loudness Normalization Works
Loudness normalization is simple in principle. The platform measures the integrated loudness of your track, compares it to its own target, and applies a gain change at playback so everything in a playlist sits at roughly the same level. It is a volume knob, applied automatically, and it does not re-encode or re-master anything.
Apple describes the mechanism plainly in its Apple Digital Masters technology brief: Sound Check measures the loudness of a track, stores that figure in the file's metadata, and then uses it to raise or lower the playback volume so songs do not lurch in level when a listener shuffles. Spotify does the equivalent with its own analysis, citing the ITU 1770 standard.
The important consequence is that the loudness you print is not the loudness anyone hears. Here is what actually happens to three different masters of the same song on a platform targeting -14 LUFS.
THREE MASTERS, ONE PLAYBACK LEVEL
Loudness normalization, as a streaming service applies it.
- Master at -8 LUFS Playback gain -6 dB Arrives at -14 LUFS, squashed to get there.
- Master at -14 LUFS Playback gain 0 dB Arrives at -14 LUFS, untouched.
- Master at -20 LUFS Playback gain +6 dB Arrives at -14 LUFS, dynamics intact.
Notice what the loud master got for its trouble: nothing. It arrives at exactly the same playback level as the moderate one, having sacrificed its transients to a limiter along the way.
The Targets, Platform by Platform
Here is where most articles list a confident table of numbers. Some of those numbers are published by the platforms and some are community measurements, and it is worth knowing which is which.
| Platform | Target | Published by the platform? |
|---|---|---|
| Spotify | -14 LUFS (Normal) | Yes. Spotify also documents Loud at -11 LUFS and Quiet at -19 LUFS, which Premium listeners can select. |
| Apple Music | Around -16 LUFS | No. Apple documents how Sound Check works but does not publish a LUFS figure. The -16 number comes from measurement. |
| YouTube | Around -14 LUFS | No. Widely measured, not documented. |
| Tidal | Around -14 LUFS | No. Widely measured. |
| Amazon Music | Around -14 LUFS | No. Widely measured. |
| Deezer | Around -15 LUFS | No. Widely measured. |
Spotify is the outlier for transparency: it publishes its target, its three listener-selectable levels, and its true peak guidance. Everyone else leaves you to infer, which is one reason the numbers vary between sources and drift over time.
The practical takeaway is that the spread across every major platform is only about 2 LU. You do not need a separate master per service, and you should be suspicious of anyone selling you one. A single master lands within a couple of decibels everywhere.
Spotify's Loud setting is the one case where a quiet master can genuinely be altered rather than just turned up. Spotify documents a limiter for that mode, engaging at -1 dB with a 5 ms attack and 100 ms decay, to stop soft dynamic tracks clipping when they are boosted. If you master extremely quietly, some listeners hear your track through that limiter rather than your own.
Why Mastering Louder Than the Target Backfires
Push a master to -8 LUFS and you have to remove roughly 6 dB of peak information to get there. That comes out of your transients: the snap of the snare, the pluck at the front of a bass note, the breath before a vocal line. Those are the details that make a mix sound big in the first place.
Then the platform turns the whole thing down by 6 dB anyway. You have paid the full price of limiting and received none of the loudness. Worse, you are now competing against a well-mastered track at the same playback level, except yours has flat transients and no dynamic contrast. It will sound smaller, not bigger. Apple makes the same point in its own documentation, noting that heavily loud masters get played back quieter and can end up sounding weaker for it.
This is the part that takes a while to believe, because in the studio, louder always wins an A/B test. Human hearing is biased toward the louder of two options regardless of quality. The moment normalization removes the level difference, that bias disappears and you are left comparing the actual audio. If you want to check your own work honestly, level-match the two versions to the same integrated LUFS before you compare them. Most people are surprised.
None of this means dynamics are automatically good. A limiter used well is a creative tool, and plenty of genres need density and sustain. The point is narrower: pushing past the target buys you nothing you keep. If you want to understand what your limiter is actually doing to the signal, our guide to understanding limiters covers the mechanics.
True Peak: The Number That Actually Causes Distortion
Loudness gets the attention, but true peak is the number that causes audible damage, and most people set it wrong.
Your DAW's peak meter reads sample peak: the highest value of the individual samples in the file. The problem is that a digital file is not the final signal. When a converter or a codec reconstructs a smooth waveform through those sample points, the curve can pass above the highest sample. That overshoot is an inter-sample peak, and true peak metering (dBTP) is what catches it.
WHY A LEGAL FILE CAN STILL CLIP
Time
Every sample is legal. The waveform between them is not.
This is not a theoretical worry. Apple ships a command line tool called afclip specifically so engineers can check the encoded file for clipping, on the basis that levels showing no overs in PCM can still clip once encoded. Lossy encoding to AAC or Ogg Vorbis shifts peaks around, and a master sitting at 0.0 dBFS has no room to absorb that.
Spotify's published guidance is to keep true peak below -1 dBTP, and below -2 dBTP if your master is louder than -14 LUFS, because a louder file gets more aggressive treatment on the way through. That second number is the one people miss.
Set your limiter's ceiling to -1.0 dBTP and make sure true peak detection is switched on. It is off by default in several popular limiters, which means the plugin is policing sample peaks and letting inter-sample peaks straight through.
So How Loud Should Your Master Be?
For music going to streaming platforms, master to around -14 LUFS integrated with a true peak ceiling of -1 dBTP. Deliver 24-bit WAV at your project's sample rate and let the distributor handle conversion.
Treat -14 as a center of gravity rather than a rule. Anywhere between roughly -13 and -16 LUFS is fine, and the difference is smaller than the difference between a good mix and a mediocre one. What matters far more is that the number lands there because the music sounds right, not because you dragged a limiter until the meter agreed.
That ordering is the whole trick. Get the mix balanced, use the limiter for the last 2 to 3 dB of control rather than 8, and then check the meter. If a mix arrives near the target without much limiting, it was a good mix. If you need heavy limiting to get there, the problem is usually in the low end or an unbalanced arrangement, not the mastering chain. Our walkthrough of mastering a track at home covers that chain in order.
A few cases genuinely differ:
- Club and DJ playout: tracks are often mastered louder, because DJ software and club systems do not normalize the way streaming does.
- Podcasts and spoken word: the usual target is -16 LUFS for stereo, and speech consistency matters more than peak loudness.
- Broadcast: follow the delivery specification you were given, typically -23 LUFS in EBU R128 territories or -24 LKFS under ATSC A/85. These are contractual, not advisory.
- Bandcamp and direct downloads: no normalization at all, so your printed loudness is what people hear.
How to Measure Your Own Track
Measuring properly takes about a minute, and the common errors all come from measuring the wrong thing.
- Put the meter last in the chain, after your limiter, on the master bus. A meter before the limiter tells you nothing about what you are exporting.
- Play the entire track from start to finish. Integrated loudness is a whole-program measurement. Measuring the chorus alone gives you a number that is 2 to 3 LU higher than reality.
- Read integrated, not momentary. The momentary reading will swing wildly and mean nothing on its own.
- Check true peak on the same pass and confirm it never exceeds your ceiling.
- Measure the exported file, not just the session. Import your bounced WAV to a fresh track and meter that, which catches export-stage surprises like a dither plugin or a stray fader.
If your integrated reading is more than about 2 LU from where you want it, fix it with the mix and the limiter's input gain rather than by yanking the output ceiling around. And if you are troubleshooting a mix that will not get loud without falling apart, the problem is usually spectral rather than dynamic, which is where a spectrum analyzer earns its place.
Genre, Dynamics, and When to Break the Rule
Loudness targets are a playback convention, not a musical value. Different genres sit naturally at different integrated levels, and forcing everything to one number is its own mistake.
| Material | Typical integrated loudness | Why |
|---|---|---|
| Modern pop, hip hop, EDM | -9 to -13 LUFS | Dense arrangements with sustained low end; commonly mastered loud despite normalization. |
| Rock and indie | -11 to -14 LUFS | Needs drum transients to survive, so tends to sit a little lower. |
| Acoustic, folk, singer-songwriter | -14 to -17 LUFS | Dynamic range is part of the performance. |
| Jazz and classical | -16 to -20 LUFS | Wide dynamics are the point; limiting them flat destroys the recording. |
Plenty of commercially successful records sit well above -14 LUFS, and their engineers know exactly what normalization will do. They accept the turn-down because the density itself is part of the sound. That is a deliberate artistic decision, and it is completely different from crushing a master because you assumed loud wins.
The rule to actually internalize: never sacrifice something you like about the mix in order to hit a number. If your track sounds best at -16 LUFS, print it at -16 LUFS. The platform will bring it up toward the target on the services that apply positive gain, and on the ones that only turn things down it will simply play a touch quieter than the playlist, which no listener has ever complained about.
Common Loudness Mistakes
Mastering to the meter instead of the music. Watching numbers while you work produces tracks that measure correctly and sound lifeless. Set the meter aside, get the balance right, then check it at the end.
Leaving the ceiling at 0.0 dBFS. This is the single most common cause of distortion that appears only after upload. Encoding shifts peaks; give it a decibel of room.
Using sample peak instead of true peak. If your limiter has a true peak option and it is off, you are not measuring what the platform measures.
Measuring only the loudest section. Integrated loudness covers the whole track, quiet intro included.
Making a separate master per platform. The entire spread is about 2 LU. One master covers all of them, and maintaining several versions mostly creates opportunities to release the wrong file.
Assuming quiet automatically means dynamic. A quiet master that was still crushed by a limiter and then turned down is the worst of both worlds. Loudness and dynamic range are separate measurements; keep an eye on LRA as well as LUFS.
Everything above assumes the mix arriving at your limiter has room to work with. If yours routinely shows up slammed against the ceiling before you have touched anything, the problem starts much earlier in the chain: see gain staging explained for the levels to hit from the microphone onward.
Get the mix right, use the limiter with restraint, land near -14 LUFS with a -1 dBTP ceiling, and stop worrying about the loudness war. It was settled by an algorithm years ago. If you want the arithmetic handled for you while you work, the dB Calculator converts between dBFS, peak, RMS and headroom, and the rest of the production tools cover the tempo and delivery maths that surrounds a master.
Frequently Asked Questions
How loud should a master be for streaming?
Master to around the platform's normalization target, which is -14 LUFS on Spotify and -16 LUFS on Apple Music. A track delivered at -11 LUFS against a -14 LUFS target simply has 3 dB of gain reduction applied on playback, so the extra loudness is discarded.
What does LUFS measure?
LUFS measures perceived loudness over time, weighted for how human hearing responds to frequency, which is why it is used instead of peak level. A 40 Hz sine and a 1 kHz sine at identical peak levels do not sound equally loud, and LUFS accounts for that difference.
What happens if I master louder than the target?
Mastering louder than the target gets your track turned down, not played louder. Three masters of the same song at -8, -14 and -20 LUFS all reach the listener at -14 LUFS on a platform targeting -14; the loud one simply arrives with less dynamic range.
What is true peak and why does it matter?
True peak measures the level of the waveform between samples, which can exceed the sample peaks a normal meter shows. It matters because lossy encoding reconstructs those in-between peaks, so a file that measures -0.1 dBFS can clip once it becomes an MP3 or AAC.