Peak normalization vs. loudness normalization
Classic peak normalization turns a file up until its loudest sample reaches a set level. It's simple, but it says nothing about how loud the audio sounds: one handclap or popped “p” can stop the rest of a quiet recording from being raised at all, and matching peak levels tell you nothing about whether two files will sound equally loud.
Loudness normalization measures perceived loudness across the whole file in LUFS, using the ITU-R BS.1770 method that streaming services and broadcasters rely on, then applies one gain change to hit a target. That's what this page does. When reaching the target would push peaks too high, a look-ahead limiter holds just those peaks under your ceiling.
Which target should you pick?
- −14 LUFS: Spotify's documented playback level, and about where YouTube's normalization lands. Louder uploads are simply turned down, so mastering hotter gains you nothing.
- −16 LUFS, the usual podcast target; Apple Podcasts asks for −16 LKFS, give or take 1 dB.
- −19 LUFS, an option some producers use for mono podcasts, which sound louder than stereo at the same reading. Enter it as a custom target.
- −23 or −24 LUFS: European (EBU R128) and US (ATSC A/85) broadcast.
Making a voice-over for a video? Use the target of the platform where the video will live. For YouTube that means about −14 LUFS: a much quieter mix can sound weak next to the videos around it.
Making a quiet recording louder
Raising a quiet recording is the most common reason to normalize. Before you commit, the panel shows exactly how much gain is coming and whether peaks will need limiting. A few decibels of limiting is inaudible; if it warns about much more, the audio will start to sound squashed. Choose a slightly lower target, or tame the loudest moments with Volume in the audio editor first.
Two things normalizing won't do. It won't balance a loud host against a quiet guest, because the whole file moves by the same amount. And it won't clean up noise: turn a recording up and its hiss and room hum come up with it.
Saving the result
Save WAV for a master or a strict delivery spec, since it can't add peaks of its own. MP3 at 320 kbps or M4A at 256 kbps is ready to upload as it is. Lossy encoding can nudge true peaks up slightly, so the result card re-measures the saved file and tells you if it crept over your ceiling. Files with more than two channels are saved as WAV to keep every channel.
Working from a video? Its soundtrack is measured and saved as audio only, so put the new file back under the picture in your editor. If only the format needs changing, the audio converter leaves the level alone; to cut or fade before normalizing, the MP3 cutter is the quickest route.