What LUFS actually measures
LUFS stands for Loudness Units relative to Full Scale. Unlike peak metering, which reports the height of the tallest sample, LUFS estimates how loud a track actually sounds to a person, using a weighting curve that follows human hearing.
This distinction matters more than it sounds. A sparse acoustic recording can peak at 0 dB and still measure -20 LUFS, because the loud moments are brief and the rest is quiet. A compressed electronic track can peak at exactly the same 0 dB and measure -7 LUFS, because it sits near its maximum almost continuously. Same peak, wildly different perceived loudness.
Three variants come up. Integrated LUFS is the average over the whole track and is the one platforms use. Short-term LUFS covers a three second window and is useful for spotting sections that jump. Momentary LUFS covers 400 milliseconds. When someone says a track is at -14 LUFS without qualifying it, they mean integrated.
The targets, platform by platform
The numbers below are the normalisation levels the platforms play back at, not rules they enforce on upload. You can upload anything; these are the levels your track will be adjusted to.
- Spotify: about -14 LUFS integrated.
- Apple Music: about -16 LUFS with Sound Check enabled, and it behaves closer to -14 in practice for most listeners.
- YouTube: about -14 LUFS, applied to the audio track of the video.
- TikTok and Instagram: roughly -13 to -14 LUFS, with short-form content tolerating slightly louder.
- Radio and club playback: -8 to -11 LUFS, which needs a separate master and should never be the file you send to streaming.
Concrete action: if you make one master, target -14 LUFS integrated with a -1 dBTP ceiling. It is the value that behaves acceptably on every streaming service, and the small differences between platforms are far less audible than the difference between a good and a bad master.
What normalisation does to your track
Loudness normalisation is simple in principle: the platform measures your integrated LUFS, compares it to its own target, and applies a gain change on playback so every track in a playlist arrives at a similar perceived level.
Crucially, the gain change is applied to the file as it is. It does not undo compression, it does not restore transients, and it does not repair anything. If you crushed your master to reach -7 LUFS, the platform turns it down to -14 and the listener hears a quiet, flat, lifeless version of your track sitting next to a competitor who mastered sensibly and kept their dynamics.
This is why the loudness war stopped paying off. Before normalisation, a louder master genuinely sounded more impressive in a playlist. Now the loudness advantage is removed automatically and only the damage remains.
If you master too loud
Mastering significantly above the target costs you twice. The gain reduction removes the loudness you fought for, and the processing used to get there stays permanently in the file.
- Transients flattened. Drums lose attack and the track stops feeling like it has impact.
- Loudness range collapses. Below about 3 LU the track sits at one level from start to finish and becomes tiring within ninety seconds.
- Distortion risk. Reaching very high loudness usually means pushing a limiter hard enough to introduce audible artefacts.
- No benefit at the other end. The listener hears it at the same level as everything else regardless.
Concrete action: if your integrated LUFS is above -12, check your loudness range before anything else. If the LRA has fallen under 3 LU, the master compression is doing damage that no amount of level adjustment will recover.
If you master too quiet
The opposite mistake is less discussed but real. Platforms do not always apply positive gain, and some listeners disable normalisation entirely. A track delivered at -20 LUFS can arrive noticeably quieter than the rest of a playlist, which reads as amateur even when the mix is excellent.
There is also a headroom argument. When a platform does turn a quiet track up, it has to check that the boost will not push peaks into distortion, and it may limit the gain it applies for safety.
Concrete action: if you land below -16 LUFS, it is usually not the limiter you need but the mix. Persistent low loudness with plenty of peak headroom generally means the arrangement has a very wide dynamic range or the balance leaves too much space between the loudest and quietest sections.
How to measure it properly
Two mistakes account for most wrong readings, and both are easy to avoid.
- Measure the whole track, not the chorus. Integrated LUFS is an average over the entire duration. Reading it over the loudest thirty seconds gives a number that is several LU too high and leads you to master too quietly.
- Measure the exported file, not the session. The file is what gets uploaded. Anything the export chain does, including a plugin that behaves differently offline, only shows up in the render.
Concrete action: render the master, then measure that file. Note integrated LUFS, loudness range and true peak together: on their own, none of the three tells you whether the master is right.
Do you need more than one master?
For most independent releases, one master is enough and the effort is better spent elsewhere. A single file at -14 LUFS with a -1 dBTP ceiling behaves correctly on every streaming service.
Separate versions become worthwhile when a specific context is important to you. A social media version at roughly -13 LUFS, with slightly more compression and reduced content below 40 Hz, survives phone speakers better. A club or DJ version pushed to -8 to -11 LUFS suits large systems. Neither should ever be the file your distributor receives.
The one thing worth checking whatever you deliver is that the numbers match your intent. Loudness, true peak and loudness range are measured values, not opinions: the same file always produces the same result, so this is the part of a release you never have to guess about.