What is LUFS, in one sentence?

A measure of how loud something sounds over its whole length, built to match human hearing, where 0 LUFS is the loudest a file can be and normal music sits somewhere between -6 and -16.

Which target should I use?

-14 LUFS for anything going to Spotify, YouTube or Apple Music, -16 for a podcast, and -23 for European broadcast. If you are matching a set of files to each other rather than a platform, any single target works as long as you use the same one throughout.

Why did my file not reach the target?

Its peaks ran out of room first. A recording with a wide gap between its average loudness and its loudest moment cannot be raised far without clipping, and clipping to hit a number would be worse than missing it.

Is this the same as compression?

No. Compression reduces the distance between the quiet and loud parts of a recording. Normalising moves the whole thing up or down by one amount and changes nothing about its internal balance.

Can I normalise several files to match each other?

Run each one through with the same target. Because the target is an absolute figure rather than a relative adjustment, files normalised to -16 LUFS will play back at the same loudness as each other whatever they started at.

Normalise Audio

Reads how loud a recording sounds using the same measurement Spotify and the EBU use, then applies the gain that brings it to your target. Peak normalisation is here too, for when that is what you actually want.

Choose an audio file or drop it hereMP3, WAV, FLAC, OGG or M4A, up to 100 MB.

Runs entirely in your browser. Your files are never uploaded.

How to use Normalise Audio

1Choose an audio file, or drag it onto the box.
2Read its current loudness and peak, which appear as soon as it is decoded.
3Pick a target, or switch to peak normalisation.
4Normalise and save the result.

Loudness is not peak level

A podcast with one door slam in it reaches 0 dBFS and sounds quiet throughout. A mastered pop track hits the same peak and sounds twice as loud. Peak level only describes the single loudest sample, which is why matching files by peak fails at the one job people want it for.

Loudness measures the whole thing the way the ear weighs it: the low end counts for less, the top end for a little more, and the figure comes out in LUFS. Streaming platforms play everything at about -14 LUFS, Apple Podcasts asks for -16, and European broadcast uses -23 under EBU R128. Matching a target means your file plays back at the same volume as everything around it.

What the measurement actually does

The file passes through two filters from ITU-R BS.1770, one that discounts anything below about 40 Hz and one that lifts the treble slightly, which together approximate what a listener hears rather than what a meter sees.

The result is then measured in 400 millisecond blocks that overlap by three quarters, and two gates throw away the quiet ones: anything below -70 LUFS absolutely, and anything more than 10 LU below the average of what is left. That second gate is what stops the pauses in a voice recording from dragging its loudness down and making the whole thing too loud once it is raised.

When the ceiling wins

Raising a quiet recording by 9 dB to reach -14 LUFS will push its peaks through the top and clip them. When that happens the gain is reduced to whatever keeps the loudest sample at your ceiling, and the page says which target it reached instead.

Getting the rest of the way needs compression or limiting, which changes how the recording sounds rather than only how loud it is. That is a decision for you to make in an editor, not something a normaliser should do quietly on your behalf.

Limits

Gain only. No compression, limiting or EQ, so a target that needs more headroom than the file has will not be reached.
Peaks are measured as sample peaks. True peak, which looks between the samples, can be a few tenths of a dB higher after lossy encoding, which is why -1 dBFS is the default ceiling.
Files shorter than 400 milliseconds cannot be measured, since that is one measurement block.
Mono and stereo are weighted as the specification says. Surround files are measured with the standard channel weights but are not otherwise handled.

Frequently asked questions

What is LUFS, in one sentence?: A measure of how loud something sounds over its whole length, built to match human hearing, where 0 LUFS is the loudest a file can be and normal music sits somewhere between -6 and -16.
Which target should I use?: -14 LUFS for anything going to Spotify, YouTube or Apple Music, -16 for a podcast, and -23 for European broadcast. If you are matching a set of files to each other rather than a platform, any single target works as long as you use the same one throughout.
Why did my file not reach the target?: Its peaks ran out of room first. A recording with a wide gap between its average loudness and its loudest moment cannot be raised far without clipping, and clipping to hit a number would be worse than missing it.
Is this the same as compression?: No. Compression reduces the distance between the quiet and loud parts of a recording. Normalising moves the whole thing up or down by one amount and changes nothing about its internal balance.
Can I normalise several files to match each other?: Run each one through with the same target. Because the target is an absolute figure rather than a relative adjustment, files normalised to -16 LUFS will play back at the same loudness as each other whatever they started at.