
Normalizing means bringing a track to a consistent, predictable loudness — so one song doesn't blast while the next whispers, and so your upload isn't automatically turned down by a streaming platform. It sounds simple, and the doing is, but there's one concept worth understanding first, because it changes which kind of normalization you actually want.
There are two ways to "normalize," and they answer different questions.
For matching perceived volume across tracks, reach for LUFS. For a simple "make it as loud as possible without clipping," peak is fine.

Streaming platforms play everything back at a reference loudness and turn louder uploads down to match, so mastering far above their target just gets undone (and often sounds worse). Spotify's loudness normalization guidance is the most explicit of the bunch: it normalizes to −14 LUFS measured to the ITU-R BS.1770 standard, and asks for true peaks below −1 dBTP. Common reference points:
| Destination | Typical target | Published? |
|---|---|---|
| Spotify | −14 LUFS | Yes |
| YouTube, Amazon | around −14 LUFS | No — measured convention |
| Apple Music (Sound Check) | around −16 LUFS | No — measured convention |
| Podcasts (spoken) | −16 LUFS (±1) | Yes |
| Broadcast (EBU R128) | −23 LUFS | Yes |
That last column matters more than it looks. Spotify states its target outright, Apple's podcast audio requirements ask for around −16 LKFS ±1 with true peak under −1 dBFS, and broadcast has a real standard behind it: EBU R 128 sets a Target Level of −23.0 LUFS. The YouTube, Amazon, and Apple Music numbers are different — they're figures engineers have measured and settled on, not specs the platforms publish. Apple's own Apple Digital Masters guide describes Sound Check at length and never names a LUFS value; it just asks you to leave 1 dB of headroom. Treat those as reliable convention that could drift, not law.
None of these are hard laws anyway — they're the loudness the platform normalizes toward. Hitting them means your track plays back at full intended level without being turned down.

Because it's measuring perceived loudness across the whole track, the result feels consistent next to other normalized audio — which is exactly the point.
Normalization sets the overall level; it doesn't shape the sound. If a track also needs EQ balance, stereo width, and gentle dynamics control to sit right, that's mastering — a fuller in-browser mastering chain handles loudness and tone together. And if you're stitching clips of different loudness into one file, normalize the pieces first, then merge them so the seams don't jump.
What LUFS should I normalize to for Spotify? Around −14 LUFS. Spotify normalizes playback toward that, so mastering much louder just gets turned back down.
Is LUFS or peak normalization better? LUFS (loudness) for matching how loud tracks feel to listeners; peak only guarantees you won't clip and doesn't match perceived volume.
Will normalizing reduce my audio quality? No — it applies gain, it doesn't re-encode. Export to a lossless format and the only change is the level.
Is normalizing the same as compression? No. Normalizing shifts the overall level; dynamic-range compression changes the difference between loud and quiet parts. See how to compress audio files for the other meaning of "compression" entirely.