Loudness Normalization Explained: Peak, RMS and LUFS

Why level is confusing

Two files can have exactly the same peak level and still sound very different in volume. A solo piano recording might hit 0 dBFS on a single loud chord but be quiet most of the time, while a heavily mastered rock song sits near its maximum constantly. To understand normalization, you need to separate three measurements.

The three ways to measure level

1. Peak level (dBFS)

The peak is the highest single sample value in the file, measured in decibels relative to full scale (dBFS). 0 dBFS is the digital ceiling; anything that would go above it is clipped, producing harsh distortion. Peak level tells you about headroom, not about how loud something sounds.

2. RMS level

RMS (root mean square) averages the signal's energy over a short window. It tracks perceived loudness better than peaks do, but it treats all frequencies equally, while human hearing is far more sensitive to the midrange than to deep bass.

3. Loudness (LUFS / LKFS)

LUFS (Loudness Units relative to Full Scale) is the modern standard, defined in ITU-R BS.1770. It weights frequencies to match human hearing and ignores very quiet passages, giving one number for how loud a whole programme feels. The US equivalent is called LKFS, and the scale is the same.

What "normalize" actually does

• Peak normalization raises (or lowers) the whole file by a fixed amount until its loudest peak reaches a target, for example −1 dBFS. The relationship between loud and quiet parts doesn't change. This is what the Normalize to −1 dB button in our Volume Booster does.
• Loudness normalization adjusts the file until its integrated loudness reaches a target, for example −16 LUFS. If that would push peaks above the ceiling, a limiter is needed.

Loudness targets you'll come across

Platforms adjust their exact values and settings from time to time, so treat these as approximate. The practical consequence is important, though: on streaming services, mastering a song extremely loud doesn't make it play louder. The platform simply turns it down, and it may sound flatter than a more dynamic master.

True peak and inter-sample peaks

When digital audio is converted back to an analogue signal, or encoded to MP3, the reconstructed waveform can briefly exceed the highest sample value. These inter-sample peaks can cause distortion even when no sample reaches 0 dBFS. That's why delivery specs ask for a true peak of −1 dB or lower, and why normalizing to −1 dBFS rather than 0 dBFS is a sensible habit.

Limiters: louder without clipping

A limiter watches the signal and turns down only the brief moments that would cross a ceiling, letting you raise the overall level further. Our Volume Booster uses a look-ahead limiter: it sees peaks a few milliseconds in advance and reduces gain smoothly, avoiding clicks. Used lightly (a few decibels), a limiter is transparent. Pushed hard, it squashes the dynamics and makes audio sound flat and tiring.

A practical workflow for quiet recordings

1. Normalize first to −1 dB, which is free and transparent.
2. Still too quiet? Add 3–6 dB more with the limiter on.
3. Listen for pumping or flatness on loud sections. If you hear it, back off.
4. Export at an appropriate bitrate with the Audio Compressor if the file also needs to be small.

Frequently asked questions