Every mastering conversation eventually lands on the same number: -14 LUFS. But most people have never actually seen what that unit measures, and they copy-paste it without understanding why. This article explains what LUFS really is, which of the four loudness numbers in your meter you should be reading, how streaming platforms use these measurements at playback, and why your true peak — not your integrated loudness — is usually the constraint that breaks a master.
What LUFS measures (and why peak meters lie)
A sample-peak meter in decibels full scale (dBFS) tells you the single highest spike in your audio file. That number says almost nothing about how loud a track sounds. A sparse, heavily transiented mix can have peaks at -6 dBFS and still sound quiet, while dense mid-range material sitting at -12 dBFS sample peak can sound crushing by comparison.
LUFS (Loudness Units Full Scale) was built to fix that mismatch. It is defined by the ITU-R BS.1770 specification (and standardized for broadcast and streaming by EBU R 128), and it approximates how a human ear weighs different frequencies when judging loudness:
- Before measuring, the signal passes through a frequency weighting ("K-weighting") that reduces the contribution of very low rumbles and extreme highs, so sub-bass thumps and hi-hat fizz do not dominate the number.
- Then the algorithm measures the energy over time and applies gating: silence is excluded from the average, and long passages are normalized by duration.
The result is a single number that correlates with perceived loudness in a way raw peaks never could. One LUFS difference is roughly one decibel of perceived level for dense material — small enough to be subtle track-to-track, which is exactly why platforms can use it to normalize your playlist.
The four numbers in your meter: know which one matters
Most loudness meters show more than one value at once. They measure different things with different time windows:
| Measurement | Time window | What it tells you | Where you'll see it |
|---|---|---|---|
| Integrated | Whole track (or selection) | The delivery target — "how loud is this record overall" | Mastering tools, and platform dashboards like Spotify for Artists after upload |
| Short-term | 3 s sliding window | Section-to-section dynamics: verse vs chorus, breakdown vs drop | Analyser plug-ins and DAW meters while you listen |
| Momentary | 400 ms | Fast changes and transients (a single kick, a vocal word) | Real-time metering; useful for spotting limiter pumping, not for targets |
| Loudness Range (LRA) | Whole-track spread | How dynamic the record is: low LRA = dense, compressed, modern loud style; high LRA = jazz, acoustic, orchestral dynamics preserved | Mastering QC reports; a diagnostic, never a target to hit |
The single most common beginner mistake is chasing momentary or short-term numbers. Integrated is your delivery number. Momentary tells you where the limiter just closed; short-term and LRA tell you whether your dynamics are alive. All four are useful — but only one of them decides how loud Spotify plays you.
What -14 LUFS and -16 LUFS actually mean
Here is the key idea most guides miss: platform LUFS targets are playback behavior, not mastering targets. When you upload a track, the platform measures its integrated loudness and applies a simple gain offset at stream time. If your master measured louder than their target, it gets turned down. If quieter — on some platforms — it gets turned up. Your file is never altered; only the volume knob moves in the player.
The current targets (verified against official platform documentation and 2025–26 mastering references):
| Platform | Integrated target | Boosts quiet masters? | True peak ceiling |
|---|---|---|---|
| Spotify | -14 LUFS | Yes | -1 dBTP |
| Apple Music (Sound Check) | -16 LUFS | Yes | -1 dBTP |
| Tidal | -14 LUFS | No | -1 dBTP |
| YouTube / YouTube Music | -14 LUFS | No | -1 dBTP |
| Amazon Music | no published target | — | -2 dBTP recommended (industry consensus) |
| Bandcamp | none (no normalization) | — | -2 dBTP recommended for headroom |
Two consequences matter in practice:
- Aim between -14 and -16 integrated if you distribute everywhere. A master at -14 plays 2 dB quieter on Apple Music than on Spotify after their offset; a master at -16 plays 2 dB louder on Spotify. Either is fine — pick one range and stay consistent across all your releases so the relative loudness of your catalog stays stable on every platform.
- The platforms that do NOT boost punish quiet masters. YouTube, Amazon and Tidal only turn things down. A dynamic -18 LUFS track will simply play quieter than its neighbors there — it never gets pulled up to -14. This is the strongest argument against delivering deliberately-quiet masters for streaming-only releases.
And yes: EDM, hip-hop and pop are still produced far louder than any of these numbers (club-oriented masters regularly sit at -9 or higher integrated). Streaming normalization erases that difference in a phone speaker — which is precisely why genre-aware mastering matters more than hitting the number: you want dynamics and transients that survive being turned down, not just a meter reading.
True peak: the ceiling that protects your master after upload
If LUFS decides how loud a platform plays you, true peak decides whether it survives the trip at all. And true peak is not what your DAW's channel meter says.
- dBFS (sample peak) — the highest individual sample in your file. At 44.1 kHz that is one of 44,100 measurements per second.
- Inter-sample peaks — when a DAC (or any codec) reconstructs the continuous waveform between samples by interpolation, the curve can swing above every single measured point. A file reading -0.1 dBFS on a sample meter can genuinely peak at +0.3 dBTP once reconstructed. Those 0.4 dB are real; they just hide between your samples.
- dBTP (true peak) — measures that reconstructed maximum, typically by oversampling the audio 4× or more and taking the highest value found. It is what a listener's hardware and the platform's encoder will actually encounter.
This matters most at encoding time. Lossy codecs (AAC, Vorbis, MP3) reconstruct signals with their own mathematics, and that process pushes inter-sample peaks even higher — a master measured at +0.3 dBTP before upload can land near +1 dBTP or more after encoding, producing exactly the symptoms people blame on "bad mastering": a hardness on loud transients, brittleness when the chorus hits, snare response that feels slightly off but is hard to name. The -1 dBTP (or -2 dBTP) ceiling in every platform spec above is headroom for this math, not vanity.
Rule of thumb: limit your true peak before anything else, then treat integrated loudness as the secondary target. A master at 0 dBFS sample peak looks fine and plays badly; a master at -1 dBTP usually survives AAC intact.
A practical workflow (measure → limit → verify)
- Measure first. Run the unmastered bounce through an EBU R 128 meter and note both integrated loudness and true peak. You cannot manage what you have not measured — "it sounds a bit quieter than my reference" is not data.
- Set your ceiling before shaping tone. Decide -1 dBTP (standard streaming delivery) or -2 dBTP (heavily limited, loud masters; Amazon-friendly). The limiter protects the file; everything else serves the music.
- Approach the integrated target with musical tools. If you are 4–6 LU below target after a clean limit, light compression plus makeup gain in the master usually gets there without pumping. Slapping +10 dB of raw gain onto a limited file only raises your noise floor and harshness along with it.
- Re-measure the rendered file — outside your DAW. Sample-peak meters on channel strips are not true-peak meters, and plug-in meters inside the session can disagree with the final bounce by more than you'd like.
- Check the platform dashboard after upload (Spotify for Artists shows measured loudness per release). The number there is what actually played — if it surprises you, compare against your step-4 measurement to find where the gap crept in.
Frequently Asked Questions
What is a good LUFS level for streaming?
For 2025–26 releases distributed to Spotify, Apple Music, Tidal and YouTube: aim for -14 to -16 LUFS integrated with true peak at or below -1 dBTP. Stay consistent across your catalog so relative loudness between your own tracks remains stable after each platform applies its gain offset.
Does a louder master sound better on streaming?
Not after normalization — platforms apply a playback gain, so a -9 LUFS master is simply turned down to match a -14 one. The difference you can actually hear comes from how loudness was achieved: a master that kept transients and dynamics clean while reaching the target sounds more expensive than one that hit the number with a crushed limiter chain.
Why do platforms require -1 dBTP instead of allowing 0 dBFS?
Because 0 dBFS sample peak does not tell you the true maximum of the reconstructed waveform — inter-sample peaks can already exceed it, and lossy encoding pushes them higher still (a +0.3 dBTP master can approach +1 dBTP after AAC). The -1 to -2 dBTP ceiling is encoder headroom: without it, loud transients come out hardened or brittle on exactly the tracks that have the least margin to spare.
Can I hear the difference between a -14 LUFS and a -16 LUFS master?
On normalizing platforms (Spotify, Apple Music), no — their playback gain equalizes perceived volume, so the same file with dynamics intact sounds equivalent at either target. The audible risk is not the 2 LU offset itself but what you did to get there: extra limiting on a dense mix costs clarity regardless of the final number. On platforms that do not boost quiet masters (YouTube, Amazon, Tidal), delivering well below -14 simply makes your track play quieter than its neighbors.
Should I hit the platform target exactly?
Within roughly ±0.5–1 LU is plenty; there is no audible or practical benefit to decimal-perfect targeting that outweighs the risk of over-limiting to reach it. What actually separates clean deliveries from broken ones: a true peak ceiling respected at render, an integrated level in the -14…-16 band for streaming, and both numbers verified on the final file rather than inside your session.
If you would rather skip the meter juggling entirely: upload your track in MP3 or WAV, pick a style from eleven genre patterns (each with its own mastering variants), and get a polished master back within minutes — free up to 3 times every day. Then judge the result on how it plays in your player of choice, not on which plugin you used to measure it.