LUFS & streaming loudness: hit your target
Streaming platforms turn every track to a similar loudness, so chasing a loud master mostly backfires. Here's what LUFS means and the numbers that actually matter.
For decades, mastering meant “make it loud.” Streaming quietly ended that. Today the number that matters isn’t peak level — it’s LUFS.
What LUFS measures
LUFS — Loudness Units relative to Full Scale — measures perceived loudness, not peak level. It’s built on the ITU-R BS.1770 / EBU R128 standard: the signal is K-weighted (filtered to roughly match how our ears weigh frequencies) and gated, so the reading tracks how loud something actually sounds, not how high its samples peak.
You’ll see three flavours:
- Integrated — the average loudness over the whole track. This is the one platforms use.
- Short-term (3 s) and Momentary (400 ms) — how loud it is right now, for spotting moments.
Why “loud” no longer wins
Every major streaming service normalizes playback to a target loudness. If your master is louder than the target, the platform simply turns it down; if it’s quieter, some turn it up. So a brick-walled, hyper-loud master doesn’t play louder than a dynamic one — it just arrives more squashed, with less punch, at the same loudness.
The platforms even collaborated on a shared recommendation (AES TD1008) to nudge everyone toward a consistent zone.
The numbers that matter
Most music platforms cluster around −14 LUFS integrated:
- Spotify, YouTube, Amazon Music, Tidal: −14 LUFS
- Apple Music: −16 LUFS
- Deezer: −15 LUFS
- EBU R128 broadcast: −23 LUFS
For peaks, keep a true-peak ceiling of −1 dBTP (dB true-peak) to leave room for inter-sample peaks and lossy encoding.
What to actually deliver
You don’t need a different master per platform. A single master at −14 LUFS integrated, true peak ≤ −1 dBTP sits right in the main cluster, and it plays fine on Apple’s quieter −16 target because Apple mostly turns content down, not up.
The practical takeaway: mix and master for the music, not the meter. Get the dynamics and tone right, land around −14 LUFS with −1 dBTP of headroom, and let normalization handle the rest. Loud for loud’s sake just costs you punch you’ll never hear back.