LUFS Loudness Meter - Integrated, True Peak and LRA

Measure loudness the way every platform does

Loudness is not level. A track can peak at exactly 0 dBFS and still sound quieter than one that peaks at -3, because how loud something SOUNDS depends on which frequencies carry the energy and on how much of the programme is actually loud rather than briefly loud. That is what LUFS measures, and it is why every streaming service, broadcaster and podcast platform now specifies a loudness target rather than a peak.

This meter reads the full EBU R128 set on your device: integrated loudness over the whole programme, loudness range, true peak found between the samples, and the momentary and short-term curves drawn over the length of the file so you can see WHERE the loudness sits rather than just what it averages to. Nothing is uploaded.

What each number means

Integrated loudness is the single figure a delivery spec is written against. It is the loudness of the whole programme, with quiet passages gated out so that silence between songs or pauses in speech do not drag the number down.

True peak is the highest point the waveform actually reaches, including between the samples. It is not the same as the sample peak your editor shows, and the difference is exactly why ceilings are specified in dBTP: a file at -0.1 dBFS by sample can breach 0 dBTP once a converter reconstructs it.

Loudness range describes how far the level travels through the programme. A dense master might sit near 3 LU and an orchestral recording near 15 - neither is wrong, but a broadcaster may specify a limit and a mix for a noisy environment usually wants less.

Momentary and short-term are the same measurement over 400 ms and 3 s windows. Their maxima are what a broadcaster checks when a spec bounds short-term loudness, and the curve is what shows you which section is the problem.

Peak to loudness (PLR) is the distance between true peak and integrated loudness - a compact way to describe how much dynamic range a master kept.

Delivery specs and streaming references are different things

A delivery specification is a contract. EBU R128 asks for -23.0 LUFS within 0.5 LU and a true peak at or below -1 dBTP; ATSC A/85 asks for -24 LUFS with a wider tolerance. A programme outside those is returned by the station, so a pass or fail is the right thing to show.

A streaming reference level is not a contract. Spotify, YouTube, Tidal and Amazon Music play everything; they simply apply gain so one track is not dramatically louder than the next. Mastering above their reference does not get you rejected and does not get you played louder - it only means the dynamics you gave up bought nothing.

So this tool shows a verdict for the first kind and a playback gain for the second. It also flags the case nobody else does: when a platform would turn a QUIET track up, and your true peak is close enough to full scale that the gain pushes it past.

How the measurement is checked

The two filter stages the standard specifies are derived at your file's own sample rate rather than applied from the 48 kHz table, because most music is 44.1 kHz and using 48 kHz constants there puts both corner frequencies about nine percent high. At 48 kHz the derivation reproduces the published table to twelve decimal places.

Accuracy is checked against signals whose loudness is arithmetic rather than against another meter. Two identical 1 kHz sine waves at a known level have a loudness that can be written down in advance; measured across levels from -6 to -40 LUFS at both 44.1 and 48 kHz, the worst error is 0.0052 LU. A sine positioned so its true peak falls exactly between two samples has a true peak of exactly 0 dBTP; it reads 0.0105 dBTP.

The same file is also measured by an independent implementation of the same standard on every test run, and the two agree within 0.04 LU on programme material.

Correcting a file to a target

Loudness is scale-invariant: multiplying every sample by the same amount moves the measured loudness by exactly that amount, whatever the material. So the correction is one gain, computed as target minus measured, and the result lands on the target rather than near it.

A limiter is inserted only when raising the file would push its true peak past the ceiling - never by default, because a limiter that runs when it is not needed changes a file for no reason and costs the correction its exactness.

Every corrected file is measured again through the same meter before it is offered to you, and the tool refuses to hand back a file whose loudness did not move by the amount that was asked for.

What it reads

MP3, WAV, FLAC, AIFF, M4A, AAC and OGG, plus the audio inside MP4, MOV, WebM and MKV - dropping a video and measuring its soundtrack is a normal thing to want, and the picture is discarded without being decoded.

The file's own sample rate and channel layout are used as they are. Surround material is weighted the way the standard specifies, with the surround channels counted higher and the LFE excluded, so a 5.1 mix reads correctly rather than being downmixed to something else first.

Anything the tool cannot read is named at the moment you choose it, with a link to the converter, rather than failing after a wait.

Questions

What is LUFS, and how is it different from dB?

LUFS measures how loud something SOUNDS, not how large its samples are. It filters the audio the way hearing weights frequency, then ignores the quiet gaps so a track with pauses is not scored as quieter than one without. Two files can peak at exactly the same level and be several LU apart in loudness, which is why every platform and broadcaster now specifies LUFS rather than peak.

Why is true peak different from the peak my editor shows?

Your editor usually shows the largest SAMPLE. The signal a converter reconstructs between those samples can go higher, and that is what clips a listener's playback. This meter reconstructs the waveform at four times the rate to find it, which is what a dBTP ceiling is defined on. A file can sit at -0.1 dBFS by sample and still breach 0 dBTP.

Will Spotify reject my track if it is louder than -14 LUFS?

No. Streaming platforms do not reject anything for loudness; they turn the file down to their own reference level and play it. The consequence of mastering louder is simply that the dynamics you gave up bought nothing. This tool shows the exact gain each platform will apply rather than a pass or fail, because a pass or fail there would be a claim about the platform that is not true.

What is loudness range, and what should mine be?

Loudness range describes how far the level moves through a programme - the difference between its quiet and loud passages, after brief extremes are discounted. There is no universally correct value: a dense electronic master might sit near 3 LU and a classical recording near 15, and both are right for what they are. It matters when a delivery spec names a limit, or when a mix will be listened to somewhere noisy.

Why does a quiet track show a warning about clipping?

Some platforms will turn a quiet track UP toward their reference level when the listener has that option on. If your true peak is already close to full scale, that gain pushes it past - and the distortion happens on playback, in a file that measured clean. The platform table flags exactly this case.

How accurate is the measurement?

On signals whose loudness is known by arithmetic, the error is at most 0.0052 LU across levels from -6 to -40 LUFS at both 44.1 and 48 kHz. The filter coefficients are derived at your file's own sample rate and match the standard's published table to twelve decimal places. An inter-sample peak that is exactly 0 dBTP by construction reads 0.0105 dBTP.

Do my files get uploaded?

No. The audio is decoded and measured on your device and nothing is sent anywhere.