Normalize audio volume
Raise a quiet recording as loud as it will go without clipping, and see by how much.
Runs in your browser. Nothing is uploaded.
A recording made too far from the microphone is quiet, and turning up the playback volume turns up the room noise with it. Normalising fixes the file instead: it finds the single loudest sample in the whole recording and multiplies everything by the amount that brings that sample up to just under full scale.
This is peak normalisation, and it is worth being precise about what that means. Every sample is multiplied by one number, so nothing is squashed and the quiet parts stay exactly as far below the loud parts as they were — the recording gets louder, not flatter. Matching the perceived loudness of several tracks is a different measurement (EBU R128) and a different tool.
The peak that was found and the gain being applied are both shown in decibels, so a file that comes back with +0.2 dB tells you it was already about as loud as it goes, rather than leaving you to wonder why it sounds the same. The headroom setting is how far below full scale to stop: a decibel by default, because a lossy encoder can overshoot the samples it was handed and a file peaking at exactly zero clips in some players.
Step by step
- 1Drop in the files.
- 2Leave the headroom at 1 dB unless you know you want otherwise.
- 3Press Normalize, then download.
Questions
Will this make quiet parts louder?
Everything goes up by the same amount, so the balance within the recording does not change. To even out loud and quiet passages you want compression, which this is not.
What is the headroom for?
It stops just below full scale. Lossy encoders can produce samples slightly louder than the ones they were given, and a file that peaks at exactly zero can clip on playback. One decibel is a safe default.
Why did my file barely change?
It was already near full scale. The panel shows the peak that was found — if it reads close to 0 dB there is nothing left to raise.