ihateaudio

Loudness Meter

LUFS, loudness range and true peak. Measured properly.

Drop an audio file here

or paste one from your clipboard

MP3 · WAV · M4A · AAC · OGG · FLAC and moreYour file stays on your device. Always.

How to use it

  1. Drop the file onto the page. Measurement starts by itself once the audio has decoded, and takes a second or two for a full track.
  2. Read the integrated loudness first. That single LUFS figure is what every streaming platform compares against its own target, and it is the number that decides whether your file gets turned down.
  3. Check the true peak. If it sits above -1 dBTP the file may clip once it is encoded to MP3 or AAC, even though it never clipped in your editor.
  4. Compare the table against the platform you are publishing to, then use the audio normalizer to move the file to that target while holding the peak below its ceiling.

What integrated loudness actually measures

Peak meters answer a question nobody asks. Knowing that the tallest sample in a file reached -0.1 dBFS tells you almost nothing about how loud the file sounds, because loudness is a function of energy over time and of which frequencies carry that energy. A bass-heavy track and a bright one can share a peak and be four decibels apart to the ear.

ITU-R BS.1770 fixes this in three stages. First it applies K-weighting. A high shelf that approximates the effect of a human head in a sound field, followed by a high-pass that discards rumble which contributes nothing to perceived loudness. Then it measures mean square energy in 400-millisecond blocks that overlap by 75 per cent. Finally it gates: blocks below -70 LUFS are dropped as silence, and blocks more than 10 LU below the average of what remains are dropped as well.

That second gate is the part people miss when they write their own meter. Without it, a track with long quiet verses reads far lower than it sounds, and normalizing to the ungated figure makes the loud sections uncomfortably loud. The gated result is what Spotify, YouTube, Apple and every podcast host compute, which is why it is what this page reports.

Two peaks, and the gap between them

Sample peak is the largest absolute value stored in the file. True peak is an estimate of the largest value the signal reaches after a converter reconstructs the continuous waveform running through those samples, and that curve can overshoot every sample it passes through. Four times oversampling, used here, catches most of the difference.

The gap matters because it is where unexplained distortion comes from. A master that shows exactly 0.0 dBFS in an editor can measure +0.9 dBTP, clip on a phone's DAC, and clip harder once an MP3 or AAC encoder has moved the waveform around. Lossy encoding does not preserve peaks; it preserves what things sound like, and the decoded output routinely overshoots the input by half a decibel or more. Every platform target therefore specifies a ceiling below zero, usually -1 dBTP, and that ceiling is insurance rather than pedantry.

Reading the comparison table

The difference column is your measured loudness minus the platform target. A positive number means the file is louder than the platform wants, and playback normalization will turn it down by that amount. Nothing is damaged when this happens. But if you compressed heavily to reach that level, you gave up dynamic range and received nothing in exchange.

A negative number means the file is quieter than the target. What happens next depends on the service. Spotify will lift quiet tracks, but only for listeners on the "Loud" setting; YouTube never lifts anything, so undershooting there costs you real loudness against everything else on the platform. Within roughly half a decibel either way, there is nothing to fix and no audible difference to gain.

Questions

What is LUFS, and why not just use dB?
LUFS is loudness weighted for how human hearing works, averaged over the whole programme, with silence gated out so pauses do not drag the reading down. A plain dB peak reading tells you the height of the tallest sample, which correlates poorly with how loud something sounds. Two files can share a 0 dBFS peak and differ by 12 LUFS.
Why does my DAW report a slightly different number?
Almost always gating or a different measurement window. This meter implements ITU-R BS.1770-4 with both gates (the absolute gate at -70 LUFS and the relative gate 10 LU below the ungated mean) which is what Spotify, YouTube and the podcast hosts use. A meter without the relative gate reads lower on material with quiet passages. Differences under about 0.3 LU are normal and inconsequential.
The true peak is higher than the sample peak. Is that a bug?
No, and it is the whole reason true peak exists. Digital audio stores discrete samples, but playback reconstructs a continuous curve through them, and that curve can rise above every sample it passes through. This meter oversamples four times to estimate the reconstructed peak. A file reading 0.0 dBFS sample peak and +0.8 dBTP true peak will clip on some hardware.
What is loudness range and what should mine be?
Loudness range, in LU, is the spread between the quiet and loud parts of the programme, measured as the 10th to 95th percentile of three-second blocks. There is no correct value, only wrong ones for a context: a modern pop master sits around 3–6 LU, a podcast around 5–8, orchestral recordings above 15. A podcast with an 18 LU range is unlistenable in a car.
Is louder better?
Not since normalization became the default. If you master four decibels above the target, the platform turns the file down four decibels on playback, and all you have done is squash the dynamics for no gain in perceived loudness. The one thing worth avoiding is landing far below the target on YouTube, which turns loud files down but never turns quiet ones up.
Does it measure the whole file or just part of it?
The whole file, which is what integrated loudness means. Blocks quieter than the gates are excluded from the average, so a long silence at the start or a fade-out at the end does not skew the result. Loudness range and both peak figures are also taken across the entire file.
My file reads as silent or shows no number.
That happens when every 400 ms block falls below the -70 LUFS absolute gate. The file is either genuinely silent or so quiet it is indistinguishable from silence. Check the waveform above the readout. If it looks flat, the recording failed rather than the measurement.