Why did normalising barely change how loud it sounds?

Because peak normalisation is driven by the single loudest moment. If your recording has one door slam near full scale, the rest stays quiet. Look at the average (RMS) figure: when the peak is high and the average is low, what you want is compression rather than normalisation.

What peak target should I use?

Around -1 dBFS for general use, which leaves a little headroom for the rounding that happens when the file is encoded to a lossy format. -3 dBFS is a safer choice if the result will be re-encoded again afterwards.

Normalise an audio file's loudness

Open the studio

Free, no sign-up, and nothing is uploaded.

Worked example

This is one of the two tools here that needs no engine download — it runs on the Web Audio API.

Engine download	none
Output format	WAV (PCM)
Runs in	Web Audio API

Because the browser already ships a decoder for MP3, AAC, FLAC, Opus and WAV, this needs no WebAssembly and no download. It is effectively instant.

Why it matters

Peak normalisation finds the single loudest sample in the file and scales everything so that peak sits just below clipping. It is the right tool for a recording that is simply too quiet. It is not loudness matching: a track with one loud transient and a quiet body will barely change, because the transient already defines the peak. The tool reports both the peak and the average level so you can see which situation you are in.

What this needs

Nothing. This runs on the browser's own audio decoder, so there is no engine download at all.

Related tools

Trim Compress Convert format Video to GIF Extract audio Remove audio Trim audio MP4 (H.264) WebM (VP9) MKV (Matroska) MOV (QuickTime)