How loud is it, and what will Spotify do to it?

No, and this page is the easy one to verify: open the network panel before you drop a track and watch it stay empty. Unlike most of our tools this one does not ask the catalog anything either — every figure it prints is computed from your samples in your tab. The file is read by the browser's own file reader, decoded by the browser's own audio engine, and measured by JavaScript that is already in the page. There is no upload endpoint behind it.

Measured rather than claimed. Run against ffmpeg's ebur128 filter — which is libebur128, the reference implementation almost every commercial loudness plugin is also built on — this meter agrees on integrated loudness to a mean of 0.012 LU, worst case 0.1 LU, across the test set. Loudness range agrees to 0.065 LU and true peak to 0.088 dB. For reference, the standard itself quotes readings to 0.1 LU, so the disagreement is below the resolution anybody uses.

The honest answer depends on where it is going, and the two answers point in opposite directions. For a club record on Beatport, nothing normalises anything and the level you master is the level that plays — which is why the 900 released records we measured sit at a median of -8.8 LUFS, with 97.9% of them louder than Spotify's target. For streaming, anything above -14 LUFS is simply turned back down, so the loudness you fought for is discarded and only the dynamics you gave up to get it remain. The genre table on this page is the useful version of the answer: the median runs from -6 in the loudest genre to -10.5 in the quietest, which is nearly 5 LU of legitimate variation.

A dB reading off a normal meter is about the sample values — how close the waveform gets to the ceiling. LUFS is about how loud it sounds over time, which is a different question with a different answer. The frequency weighting is most of the difference: the same peak level made of sub bass and made of hi-hats measures the same in dBFS and very differently in LUFS. The practical consequence is that peak level tells you whether a file will distort and tells you nothing at all about whether it will sound quiet next to the record before it.

Because your peak meter is reading the stored samples and the speaker plays the curve between them. Reconstructing that curve — which is what a converter and a lossy decoder both do — routinely produces a value above the highest stored sample. That is why the delivery convention is -1 dBTP rather than -0.1: the margin exists to absorb it, and we measured how much it has to absorb. Taking 49 records, limiting each to exactly -1.0 dBTP and then encoding, a 320 kbps MP3 moves the peak by a median of nothing at all and never crosses full scale. A 128 kbps AAC — which is what a streaming service serves on a mobile connection — moves it by a median of 2.3 dB and ends up over 0 dBTP on 89.8% of them. Spotify asks for -2 dBTP on loud masters for exactly this reason.

Yes, and it is one of the reasons this exists. Files up to 800 MB, which is roughly 75 minutes of 24-bit stereo WAV, and the meter is written so that an hour of audio costs the same memory as a single track — the loudness accumulates as it goes rather than being held all at once. The loudness-over-time curve on a full set is the useful picture: it shows you the arc you actually played rather than the one you remember.

Loudness range is the spread between the quiet and loud parts, in LU, ignoring the extremes at both ends. It is the number that says whether a set has an arc or is one flat wall. Across the records we measured the median is 5.1 LU, with the middle half between 3.3 and 7.4 — and that is measured over a two-minute excerpt, so a full track's range is wider than those figures suggest. There is no correct value: a peak-time tool with a deliberately flat energy profile and a broken-beat record with a long breakdown are both doing their job.

Because they are solving different problems. A streaming service plays your record next to a hundred thousand others and its job is to stop the listener reaching for the volume, so it normalises everything to one level. A store sells you a file to play yourself, so it hands it over untouched. This is the single most useful thing to understand about mastering for both: the club master and the streaming master are the same file, and the only thing loudness buys you is the one place nothing turns it down.

For checking a bounce, yes — it is the same standard and it agrees with the reference implementation to under a tenth of a unit. For mastering, no, because a meter you can only run on a finished file cannot show you a change as you make it. The useful pairing is to work with whatever is in your DAW and check the export here, particularly the true peak, which is the figure most in-DAW meters get wrong by not oversampling.