Which mode should I choose by default?

Peak, if the file is going somewhere that must not clip and there is no loudness specification to meet. Loudness, if several files have to sit together at a similar perceived level, which is the case for a set of recordings made at different times.

Why did the loudness target not get reached?

Because the ceiling blocked it. A file with one loud transient and a low average needs a lot of gain to lift its average, and that much gain would push the transient past the ceiling. The tool takes the smaller gain and reports the level it actually achieved instead of hiding the shortfall.

Does normalizing change the character of the audio?

Not in peak mode, where every sample is multiplied by the same factor and the balance between loud and quiet is untouched. In loudness mode the same is true, since one gain is computed for the whole file; what changes is only the overall level.

Why is one gain used for both channels?

Because a stereo file has a balance that depends on the relative levels of the two sides. Computing a gain per channel would move a centred element to one side whenever one channel happened to be louder, so the measurement is taken jointly and one gain is applied to both.

Does it make a quiet recording as loud as a commercial one?

Only up to the ceiling. Raising the average without raising the peak requires reducing the dynamic range, which is compression rather than normalization. A compressor run before this tool is the usual way to close the remaining gap.

Audio Normalizer

You Might Also Need

Audio Amplifier

Increase audio volume safely

Loudness Normalizer

Normalize to LUFS loudness standards

RMS Normalizer

Normalize audio using RMS levels

Peak Limiter

Prevent audio peaks and clipping

Audio Normalizer brings a file to a chosen level, either by its loudest sample or by its average loudness. The two modes answer different questions, and the panel reports which one applied.

What is Audio Normalizer?

Pick a mode, set the target, drop a file, and process. The report states the level before and after, and in the loudness mode it tells you honestly when the ceiling stopped the target being reached.

The controls:

Two mode buttons, one labelled for peak and one for average loudness.
In peak mode, a target from twelve decibels below full scale to full scale, in half-decibel steps, defaulting to minus one.
In loudness mode, a target from minus thirty to minus nine decibels, defaulting to minus eighteen, plus a ceiling from minus six to zero, defaulting to minus one.
A report line giving the measured input level and the level actually achieved, with an explicit note when the ceiling got in the way.

Peak normalization measures the largest sample in the file and multiplies everything by whatever gain puts that sample on the target. Because the peak is the largest value present, no gain chosen this way can push anything past full scale, which makes the mode safe by construction and exact by definition. Loudness normalization measures instead the square root of the mean of the squares, which is a fair approximation of how loud the material sounds rather than how loud its worst moment is. That measurement is taken jointly across the channels rather than per channel, and a single gain is derived and applied to all of them, which is the detail that preserves stereo balance: normalizing the channels separately would move a centred element off-centre as soon as one side happened to peak higher. The gain is then limited by the ceiling. If reaching the loudness target would push the loudest sample above the ceiling, the smaller gain wins, and the report says so plainly rather than quietly clipping. A file with a high peak and a low average, which is what a sparse recording with one loud transient looks like, is exactly the case where the target will not be met. Both modes are two-pass operations: measure first, then scale, so nothing depends on the order the channels are visited.

What it does not do:

No dynamic processing, so the relationship between loud and quiet passages is unchanged.
Loudness mode will miss a low target on peaky material, and says so rather than clipping.
No true-peak measurement, so inter-sample peaks are not considered.
No measurement over time, so there is no loudness range or short-term view.
Peak mode gives no indication of perceived loudness at all, only of the sample maximum.
Output is sixteen bit WAV.

Use it for the two jobs the modes describe: hitting an exact peak so a file cannot clip downstream, or bringing a set of recordings to a consistent perceived level so they sit together. When the goal is a broadcast-standard loudness measurement rather than a quick level match, the loudness normalizer follows the same intent with the full standard behind it, and when the level simply needs raising without a target in mind, the amplifier adds a fixed amount of gain and warns you about clipping.

How to use Audio Normalizer

Drop an audio file onto the panel and let it decode.
Choose peak if an exact maximum matters, or loudness if perceived level matters.
Set the target, and in loudness mode also the ceiling that must not be crossed.
Press process, read the achieved level in the report, and download the result.

When to use Audio Normalizer vs related tools

Audio Normalizer is the quick level match, whether the goal is a safe peak or a consistent average. Loudness Normalizer answers the same question against a published standard, and RMS Normalizer is the narrow version that measures only average level. When the level needs a fixed lift rather than a target, Audio Amplifier and Volume Booster add gain, the second with a limiter behind it.

Privacy & Security

This tool runs entirely in your browser — no data ever leaves your device. There is no server round-trip, no upload, no logging, and no account required. Your input is processed locally using client-side JavaScript and is never stored, transmitted, or accessible to anyone else. When you close the tab, everything disappears.