Understanding Peak and RMS Volume Normalization
What this tool does
Recorded audio rarely arrives at a consistent volume. One clip peaks near the ceiling while the next sits ten decibels lower, and listeners are left reaching for the volume control. This tool measures your file and then rewrites its gain so the level lands where you want it.
When you drop in a track, it decodes the audio and reports two numbers in decibels: the max volume (the single loudest sample, called the peak) and the average volume (the RMS level, the average strength of the whole signal), alongside the duration and channel count. Those measurements are what each adjustment mode acts on.
Three ways to adjust the level
Maximize Volume applies peak normalization: it raises the whole signal until the loudest sample reaches your target, which is the safest way to make a quiet recording as loud as possible without clipping. Balance Volume works on the average volume (RMS) instead, moving it toward a consistent target so several files play back at a similar volume rather than matching only their peaks. Because it aims at the average, the loudest parts can end up above the ceiling — the tool tells you when that happens.
Custom Adjustment skips the measurement targets and shifts gain by an exact number of decibels — useful when you already know a file needs, say, +3 dB or a small cut to sit better in a mix. Levelling is usually one link in a short chain: the audio trimmer clears the silence and false starts at each end first, and a recording brought to an even level is also easier for speech to text to transcribe.
Example: leveling three interview clips
Say you have three interview recordings: one loud, one faint, and one in between. Played back to back they are jarring. Choose Balance Volume, aim for a common average target, and each clip is nudged toward the same average volume — the quiet one comes up, the loud one eases back. If a clip's loudest parts would go past 0 dB, the result says so after the run and offers to redo it in Maximize Volume mode, which keeps the loudest point at or below your target.
Notes and edge cases
Boosting gain also lifts the noise floor, so a hissy recording sounds hissier once it is boosted — normalization changes level, not the signal-to-noise ratio. Anything that would go past 0 dB is cut off flat at 0 dB (this is a hard cut, not a limiter), which can sound harsh; the tool counts those samples and warns you after the run. Play the result before you download.
Input accepts MP3, WAV, OGG, AAC, M4A, FLAC, and WebM up to 500MB; output is MP3 or WAV, with WAV staying lossless for further editing. The whole process runs locally in your browser, which means very long files depend on how much memory the browser can spare.