Change audio volume

How to Apply the volume change

1. 1

   Add your audio file

   MP3, WAV, FLAC, M4A, AAC, OGG or Opus, up to 200 MB per file and twenty at a time; each file is measured on its own.

2. 2

   Pick a mode and a target

   A fixed gain in decibels, a peak target, or a loudness target in LUFS. The measured values are shown with the result.

3. 3

   Click Apply and download

   The result is a WAV, with the actual change in decibels reported next to it.

Peak normalisation and loudness normalisation are not the same thing

Peak normalisation finds the single loudest sample in the file and scales everything so that sample lands on the target, usually just under full scale. It is fast, it guarantees nothing clips, and on many files it does almost nothing useful. A quiet recording with one door slam in it is already at the ceiling because of the slam, so peak normalisation leaves it quiet. Loudness normalisation measures how loud the file sounds across its whole length and scales to that, which is why every streaming service uses it and why two files normalised this way match when played one after another. Use peak when you need a hard guarantee about the maximum sample; use loudness when you want files to sound equally loud.

What LUFS means and why −16 is the default here

LUFS stands for loudness units relative to full scale, and it is defined by the EBU R 128 standard. The measurement applies a filter shaped like human hearing, takes the energy in 400-millisecond windows, and then throws away the quiet windows so that pauses between words do not drag a speech recording's number down. That gating is what makes the figure match perception instead of arithmetic. Streaming platforms target somewhere between −14 and −16 LUFS; podcast distributors commonly ask for −16 mono or −16 to −19 stereo. The default here is −16 because it suits spoken word without needing a limiter to work hard, and because a file at −16 survives being turned up more gracefully than one pushed to −14.

What the limiter is for

Multiplying every sample by the same number is the whole of a gain change, and if the number is bigger than one, some samples will try to go past full scale. Without a limiter those samples are cut flat at the ceiling, and a flat top on a waveform is a sudden change in slope, which the ear hears as a click or a crunch, not as loudness. The limiter compresses everything approaching the ceiling along a smooth curve instead, so the loudest moments end up very slightly quieter than a straight multiplication would make them, and nothing acquires an edge. It is on by default for gain increases and does nothing at all when you are turning a file down, where there is no ceiling to hit.

The numbers are measured, not assumed

Peak before and after, the loudness reading when it was needed, and the gain that was actually applied are all reported with the result. When there is nothing to measure, because the file is silent or shorter than the 400-millisecond window the loudness standard needs, the tool refuses rather than inventing a figure. A guessed measurement would produce a guessed gain, and on a quiet file that means either silence or a blast.

Frequently asked questions