21 April 2026 | 6 min read
Normalising audio: peak, RMS and loudness, and which one you need
Normalising to 0 dBFS is the default in most tools and almost never the right choice for professional delivery.
Normalising means applying a single gain change so the file reaches a target level. It is not compression, it does not change dynamics, and it is not a substitute for mixing. What varies is how the target is measured.
Peak normalisation
The tool finds the loudest single sample and scales the whole file so that peak lands on your target, typically -1 or -3 dBFS. It is fast and safe against clipping, but says nothing about perceived loudness: a sparse dialogue track and a dense music bed peaking at the same value will sound very different.
RMS and loudness
RMS averages energy over time, which correlates better with what you hear. Loudness measured in LUFS goes further by weighting frequencies the way human hearing does, and it is the basis of every modern broadcast spec.
- -23 LUFS: EBU R128, European broadcast.
- -24 LUFS: ATSC A/85, North American broadcast.
- -16 to -14 LUFS: podcasts and streaming platforms.
What to do with production sound
Do not normalise raw location recordings before editing. You destroy the level relationship between takes, and a single loud door slam can pull an entire dialogue track down. Keep the original gain structure, edit, then set loudness on the finished mix.
Headroom before conversion
If you are exporting to MP3 or AAC, leave at least 1 dB of headroom. Lossy encoders can produce peaks slightly above the original, and a file normalised to exactly 0 dBFS will clip on playback.