Targets: Spotify/YouTube -14 LUFS, Apple Music -16 LUFS, EU/AU broadcast (R128) -23 LUFS (+/-1 LU).
True Peak ceiling: streaming and broadcast -1 dBTP. dBTP catches inter-sample peaks a normal meter misses.
LRA = Loudness Range (in LU): quiet-to-loud spread; pop ~5-8 LU, film ~15-20 LU.
Gating (BS.1770): ignores gaps below -70 LUFS absolute and a -10 LU relative gate when averaging.
Two songs at the same -1 dBFS peak can differ 6+ LUFS in felt loudness; +10 LU ≈ 'twice as loud'.
How it works
Pass the audio through a K-weighting filter so it's weighted like the ear hears.
Square the signal and average the energy over the chosen window (400 ms / 3 s / whole track).
Apply gating: drop near-silent gaps so pauses don't fake a quiet reading.
Convert that mean energy to a dB-style value, anchored so 0 = full scale.
Compare your Integrated number to the platform target (e.g. -14 LUFS) and adjust gain to match.
Check True Peak stays under -1 dBTP so normalisation/limiting won't clip.
Real examples
Mix a podcast to -16 LUFS Integrated so speech sits at a consistent level across episodes.
Deliver a TV ad at -23 LUFS (R128) or the broadcaster rejects it for being too hot.
Two tracks both peak at -1 dBFS but read -8 vs -14 LUFS: the -8 feels far louder.
Master a single to -14 LUFS so Spotify plays it at full level instead of turning it down.
A track squashed to -6 LUFS gets pulled DOWN to -14 on Spotify, losing its 'loud' edge.
How it helps in live sound
Set console/playback meters to LUFS (Momentary + Short-term) for walk-in/interval music: -16 to -18 LUFS for comfy background, -12 to -14 for energy.
Loudness-match all backing tracks and stings to the SAME Integrated LUFS so cue-to-cue level is even and you're not riding the fader.
Watch Short-term (3 s) LUFS during speeches to keep a presenter on a steady target instead of chasing peaks.
Keep file playback under -1 dBTP True Peak so cheap media players and processors don't inter-sample clip.
Pre-normalise every track in the show folder to one LUFS target (Reaper / iZotope RX / Auphonic) so nothing jumps.
Judge 'too quiet / too loud overall' by Integrated LUFS, not the dBFS peak meter.
Everyday analogy
Peak metering is checking the single tallest wave that hit the beach; LUFS is asking how wet the whole beach got over the afternoon.
Watch out
Myth: a louder peak means a louder mix. Wrong, peak is one instant; LUFS measures sustained perceived loudness, so two -1 dBFS files can differ 6+ LU.
Fun fact
The -0.691 constant baked into the LUFS formula exists purely to calibrate the K-weighting filter so a 1 kHz tone reads the same on the new scale as the old one.
Key takeaways
LUFS = perceived loudness over time, measured the way ears hear, not split-second peaks.
0 LUFS = full scale; real music lives around -9 to -23 LUFS (always negative).
Targets: Spotify/YouTube -14, Apple -16, EU/AU broadcast -23 LUFS.