Make audio louder, or normalize it
Measure how loud a file actually is, then raise it by a known amount instead of guessing. Choose a target peak for normalizing, or add a fixed number of decibels. A limiter keeps the peaks from clipping.
Peak is not loudness
Two files can both report a peak of -0.1 dBFS and sound completely different. A single door slam sets the peak; the speech underneath can be 25 dB quieter. Normalizing to the peak then does almost nothing for the words, because the loudest sample is already at the ceiling and there is no headroom left to give.
The measurement that tracks how loud a file sounds is RMS, the root mean square of the samples — effectively the average energy. The gap between peak and RMS is the crest factor, and it is the number that tells you whether a file is "quiet but clean" or "already as loud as it can be".
This page reports peak, RMS and crest factor in dBFS before you change anything, and again after export. That is the difference between guessing at a volume slider and knowing what you did.
How is the gain calculated in normalize mode?
Target amplitude is 10^(target/20). Measured peak amplitude is the largest absolute sample in the file. Gain in dB is 20 × log10(target amplitude ÷ peak amplitude). For a file peaking at 0.25 (about -12 dBFS) with a target of -1 dBFS: 20 × log10(0.891 ÷ 0.25) = 20 × log10(3.565) = 11.0 dB. The page prints this arithmetic with your own numbers rather than hiding it.
Why does the file still sound quiet after normalizing?
Because the peak was set by one transient — a click, a door, a plosive — and the speech sits far below it. Normalizing protects that transient and leaves the speech where it was. The fix is limiting the transient first, or compressing the file so the dynamic range comes down; neither is a pure gain operation. Cutting the offending transient with the Audio Trimmer is often the fastest route.
What does the clip counter mean?
It counts samples sitting at exactly full scale in the source, which usually means the recording overloaded at the input. Those samples are already destroyed — the waveform was flattened — and no amount of gain brings them back. If the counter is high, record again with more headroom rather than processing this file.
Is my file uploaded anywhere?
No. Everything runs in the tab. Nothing is sent and nothing is stored.