Peak vs Loudness Normalization
Peak normalization scales audio so its single loudest sample hits a target (say -1 dB). It prevents clipping but says little about perceived volume - a track with one loud snare hit and a whispered verse still sounds quiet overall. Loudness normalization instead measures average perceived loudness in LUFS (Loudness Units relative to Full Scale) across the whole file and scales to a target. This matches human hearing, which is why broadcasting and streaming standardised on it.
The LUFS Targets That Matter
Two numbers cover most real-world work: -14 LUFS, the reference level used by major music streaming platforms (Spotify, YouTube Music, Amazon Music all normalise playback around it), and -16 LUFS, the widely recommended target for podcasts (Apple's guidance; many producers use -16 for stereo and -19 for mono). Broadcast television uses -23 or -24 LUFS depending on region. Delivering close to the platform target means your audio is not turned down (or noisily turned up) on playback.
Step by Step: Normalizing on mp3.now
- Open the MP3 normalizer.
- Upload your MP3 (10 MB as a guest, 50 MB with a free account).
- Review the options, then click Convert - the file's loudness is analysed and levelled to a consistent target.
- Download the normalized MP3 and spot-check the quietest and loudest sections.
Normalizing multiple episodes or an album? Process up to 10 files in a batch so everything lands at the same level.
When to Normalize - and When Not To
Normalize before merging segments recorded at different levels, before publishing a podcast episode, and after extracting quiet audio from video. Two limits to keep in mind: normalization scales the whole file, so it will not fix a recording where one speaker is loud and another quiet within the same file (that needs compression or per-track editing at the source). And it raises noise along with signal - a hissy recording normalized louder is a louder hissy recording.
Normalization vs Compression vs Limiting
Three level-related terms get mixed up constantly. Normalization applies one fixed gain change to the entire file - the dynamics between loud and quiet moments are untouched, everything just moves together. Dynamic range compression continuously turns down the loud parts (and often brings up the quiet ones), actively reshaping dynamics - this is what fixes a recording where the speaker drifts toward and away from the microphone. Limiting is extreme compression applied only at the very top, guaranteeing peaks never exceed a ceiling. A polished podcast chain typically uses all three at production time: compress for evenness, limit for safety, then normalize to the LUFS target. Note also that dynamic range compression has nothing to do with data compression - shrinking an MP3's file size changes how the audio is stored, not how loud it is.