LUFS meter for a mix or a master
True peak, LUFS loudness and the penalty each platform will apply, dynamics, mono compatibility, energy balance and narrow resonances found to the nearest tenth of a hertz. On a file, or live on one of your audio interface inputs.
1
Analyze a file
Drop a mix or a master: you get what holds, what does not, and what to try.
Analyze a file
Drop a mix or a master: you get what holds, what does not, and what to try.
Drop a mix or a master here
or
WAV, MP3, M4A, FLAC — the file stays on your computer, nothing is sent.
Very short impacts, the saturation you deliberately look for on a kick. Nothing to fix if it is intentional.
A 4.6 dB penalty on delivery. Common and deliberate in urban music; beyond six, the dynamics you gave up are nowhere to be found.
A narrow peak, present 84 % of the time, carrying 31 % of the low-band energy on its own. A narrow EQ cut here drops the measured energy without the ear hearing a difference.
2
Meter live, without exporting
Send a track or a bus to a spare output, loop it back to an input, and watch while it plays.
Meter live, without exporting
Send a track or a bus to a spare output, loop it back to an input, and watch while it plays.
Two ways in, and they do not behave the same.
An input. Send your master to a spare output and loop it back to an input — a cable, or your interface's own loopback. This works on every machine and every browser, and it is what a two-in interface gives you: one device, two channels, taken as left and right. Point it at a microphone instead and you are measuring the room, not the mix.
What the computer is playing. No browser can tap a sound card's output directly — that is an operating-system privilege, not a missing feature. The one door that exists is audio sharing: on Windows and Linux it hands over everything the machine plays, your DAW included; on macOS it only ever gives the sound of a tab. Try it and the page tells you what you actually got.
Live gives you sample peak and momentary loudness (a 400 ms window). True peak and integrated loudness cannot exist in real time — integrated is gated across a whole track, true peak needs oversampling. For the numbers you deliver on, drop the bounced file into the analysis above.
Highest peak of the last 3 seconds: −∞ dBFS
Input names only appear once permission has been granted — that is a browser protection, not a fault. Nothing is recorded or transmitted: the signal is measured on the fly, then forgotten.
What these numbers mean
True peak is the level to check before delivery, because it catches what a sample-peak reading cannot. Between two samples, the wave a converter reconstructs can rise higher than either of them: a file displayed at 0.0 dBFS can therefore overshoot at the output, and distort once encoded to MP3 or AAC. That is why platforms specify headroom in dBTP rather than dBFS.
Perceived loudness, in LUFS, is what platforms measure to bring every track to the same volume. Above their target, they turn yours down. That is not a fault for all that: nearly every current urban master deliberately accepts a penalty of a few decibels, because a track has to sound like the ones it will be heard next to. What matters is knowing how much you give up — and having decided it.
Mono compatibility is checked because a large share of listening happens on a single speaker: phone, laptop, bar speaker, story. Whatever cancels in mono disappears for those listeners, and it is usually the bass or the vocal.
Resonances are the only complaint this tool allows itself to make about your spectrum, because they are the only one that does not depend on style. Saying “too much low end” means nothing when the huge low end is the genre. But a narrow peak a few hertz wide, standing above its neighbors and staying there from beginning to end, is a fault everywhere — often the room, often speakers that never showed it. Correcting it drops the measured energy without taking anything away from what you hear, and that is exactly what a broad EQ move cannot do.
What each platform turns your track down to
Streaming services play everything at a comparable volume. They measure your integrated loudness, compare it to their own target, and apply the difference on playback. Nothing is re-encoded and nothing is destroyed — the track is simply played quieter. These are the targets the tool measures against:
| Platform | Target | Source | A −8 LUFS master is played |
|---|---|---|---|
| Spotify | −14 LUFS | Published by the platform | 6 dB quieter |
| YouTube | −14 LUFS | Measured | 6 dB quieter |
| Tidal | −14 LUFS | Measured | 6 dB quieter |
| Amazon Music | −14 LUFS | Measured | 6 dB quieter |
| Deezer | −15 LUFS | Measured | 7 dB quieter |
| Apple Music | −16 LUFS | Measured | 8 dB quieter |
| Broadcast (EBU R128) | −23 LUFS | Published standard | 15 dB quieter |
Spotify documents its target; the rest are measured values. Spotify states −14 LUFS in its own artist documentation, and EBU R128 is a published broadcast standard. The others are the figures the industry measures in practice rather than numbers the platforms publish, and the distinction matters if you are going to plan a master around them. Broadcast sits in the table as a landmark rather than a destination: urban music is not mastered to −23 LUFS.
A penalty is not a failure. Almost every current master accepts a few decibels, on purpose, because a record has to sit next to the ones it will be played beside. What is worth knowing is how many decibels you are giving up, and that you chose it.
The tool situates. It does not judge.
A track with a hundred million streams can show thousands of clipped samples and a penalty of several decibels. Measuring that is easy; telling its author it “does not comply” is useless. In urban music nobody meets the broadcast targets — and nobody misses them at random either. So the standard is stated as a landmark, never as a sentence, and no verdict here holds it against a record that it is not a classical recording.
Two exceptions are strict on purpose, and they are not matters of taste: mono compatibility, because what cancels really does disappear, in every genre; and resonances, because a narrow peak is a fault everywhere. Style ranges will come from measurements on real records — until they are measured, they are not invented here.
Frequently asked
Is my file uploaded to your servers?
No. The track is decoded by your browser and analyzed on your own machine. Nothing is sent, nothing is kept — cut your connection once the page has loaded and the tool still works.
How do I meter live, without exporting?
Send the track or the bus you care about to a spare output on your audio interface, loop that output back to an input — with a cable, or with your interface's loopback function if it has one — then pick that input from the list. The analysis runs on what comes in, in real time. You can keep your normal monitoring alongside: it is one more output, not one instead.
Why true peak, and not simply the sample maximum?
Between two samples, the wave a converter reconstructs can rise higher than either of them. A file reading 0.0 dBFS can therefore overshoot once converted, and distort after MP3 or AAC encoding. It is the true peak (dBTP) that platforms look at, not the sample level.
Is loudness measured the way platforms measure it?
Yes: LUFS to the ITU-R BS.1770 standard, with K-weighting and the double gate. The measurement was checked against the EBU reference signal — it lands within 0.01 LU. That is what makes it possible to state, without guessing, how much Spotify or Apple Music will turn your track down.
What LUFS should I master to for Spotify?
−14 LUFS integrated is the level Spotify plays everything back at, with a true peak at or below −1 dBTP. Master louder and Spotify simply turns the track down by the difference; master quieter and it is turned up, up to its limit. Neither is a fault. What you are really choosing is the dynamic range you keep: a master at −8 LUFS is played 6 dB quieter than it left your room, so all the loudness you fought for is given back, and only the crushed transients remain.
What is the loudness penalty, exactly?
The difference between your integrated loudness and the platform's target, applied on playback. A track at −8 LUFS on Spotify (−14) carries a 6 dB penalty. Nothing is re-encoded and nothing is damaged — it is a volume change. The reason it matters is that it undoes the loudness war: two tracks at −8 and −14 LUFS reach the listener at the same volume, and the quieter master is the one that kept its punch.
Do all the platforms use the same target?
Almost. Spotify, YouTube, Tidal and Amazon normalize to −14 LUFS, Deezer to −15, Apple Music to −16. Broadcast is a different world at −23 LUFS (EBU R128). The spread between streaming services is 2 dB, which is small enough that one master serves them all — you do not need a version per platform.
Why will you not tell me whether my spectral balance is good?
Because there is no such thing in the absolute. A trap record, a folk song and a string quartet do not share a curve, and none of the three is wrong. Handing you a target number would be inventing a proof. The distribution is shown so you can compare your own versions with each other, or with a reference track you know well. What the tool does flag is resonances: those can be diagnosed without knowing anything about your style.
What is a resonance, and why does fixing one change everything?
A loaded band is not a fault — in urban music the low end is huge because the style wants it that way. A resonance is something else: a narrow peak, a few hertz wide, standing above its immediate neighbors and staying there from beginning to end. Often the recording room, or a monitoring system that never showed it. The tool gives you its frequency to the tenth of a hertz and the share of energy it carries on its own. A narrow EQ cut at that exact spot brings the measured energy down without the ear hearing a difference — where a broad EQ move would cost the track its character.
How do you know it is not simply a bass note?
That is the real trap, and it is handled: a bass hammering a low E produces exactly the same peak as a room mode. What separates them is time — a note comes and goes, a resonance stays. So the tool measures the persistence of the peak across the whole duration and only shows what holds. The rate is printed next to each resonance: you decide whether a peak present 60 % of the time is a fault or your bass line.
Why will you not just tell me my master is bad?
Because a verdict would have to come from a target, and there is no honest target for most of this. A −8 LUFS master is a mistake on a jazz record and the correct decision on a drill record; the number is the same, the answer is not. What the tool can tell you without inventing anything is the measurement, what each platform will do with it, and how far you are from a reference you chose. The judgement is the part you are qualified for and it is not.
Can the tool fix what it finds?
No, and that is deliberate: a diagnosis is not a treatment. It tells you where to look and what to try. Everything happens in your browser, so nothing is processed, nothing is stored and nothing is charged for.