Peak or Loudness: How to Normalize Audio Volume

Last updated: November 12, 2026

Normalizing audio means changing the volume of a recording by one fixed amount to reach a target. It does not squeeze or reshape the sound. It simply makes the whole file louder or quieter. There are two common targets.

Peak normalization

The loudest point of the file is set to a chosen level, often a little below the maximum, such as -1 dB. Everything else moves up or down by the same amount. This is useful when you want the recording as loud as possible without clipping. But two files with the same peak can still sound very different in loudness, because a single loud cough can set the peak.

Loudness normalization

The average level of the speech is set to a target, so different files sound equally loud to the listener. For speech, an average around -20 dB is a common, comfortable choice, and -16 dB is louder. This is better when you deliver many files that should sound consistent.

The limit that protects you

If a file has big peaks, raising the average to the target could push those peaks past the maximum and cause clipping. The Sayvoi Audio Cleaner never lets the loudest point go above -1 dB, so in that case the file ends a little quieter than the target. That is expected and keeps the sound clean. To understand clipping, read audio clipping and distortion.

Which one should you choose?

• One recording that should be as loud as possible: peak.
• Many recordings that should sound alike, such as a voice data project: average loudness.
• Always follow the project guide if it names a level.

Normalize last

Noise reduction and silence removal change the level of a file. So do them first and normalize at the end. See the full order in how to clean audio recordings.

Normalizing cannot fix a recording that is already distorted or very noisy. For that, record again in a quieter place, as explained in how to record clean audio with your phone.

All articles