What LUFS to master for: streaming loudness targets

What LUFS actually measures

LUFS stands for Loudness Units relative to Full Scale. It is a perceptual measurement defined by the ITU-R BS.1770 algorithm and the EBU R128 recommendation, designed to track how loud audio actually sounds rather than its raw peak or RMS level.

The key number is integrated loudness: a single LUFS value averaged across the whole program, weighted for how human hearing responds across the frequency range. Two masters can hit the same peak level and still differ by several LUFS, which is why peak meters alone never tell you whether a track will sit right on streaming.

LUFS, dBFS and dBTP: three meters not to confuse

Reading a loudness meter correctly prevents most delivery mistakes. Three measurements look similar and mean different things.

• LUFS: perceived loudness across the program, K-weighted. This is what platforms normalize.
• dBFS: the raw digital sample peak, a purely technical ceiling with no perception weighting.
• dBTP (true peak): the real inter-sample peak reconstructed after conversion, the one that clips lossy codecs.
• Integrated, short-term (3 s) and momentary (400 ms) are the three LUFS time windows, with gating at -70 LUFS absolute and -10 LU relative.

Streaming loudness targets

Streaming services normalize playback to a reference loudness so listeners do not have to ride the volume between tracks. Mastering close to the target keeps you in control of how the adjustment sounds.

• Spotify: -14 LUFS integrated, true peak -1 dBTP (-2 dBTP if your master is louder than -14 LUFS).
• Apple Music (Sound Check): around -16 LUFS integrated.
• YouTube / YouTube Music: around -14 LUFS integrated.
• Tidal: around -14 LUFS integrated.
• Amazon Music: around -14 LUFS integrated, true peak -2 dBTP.
• Deezer: around -15 LUFS integrated.
• SoundCloud: no loudness normalization.

How normalization really works (turned up vs turned down)

Platforms measure your integrated loudness and apply a gain change to reach their reference. The direction matters and is where most confusion lives. Spotify and Apple Music can turn a too-quiet master up, but only within the true-peak limit. YouTube, Tidal and Amazon only turn loud masters down, never up. One caveat for Apple Music: its Sound Check normalization is off by default, so Apple only adjusts playback level for listeners who have switched it on.

A worked example: a master at -20 LUFS with peaks at -5 dBFS asks Spotify for +6 dB to reach -14 LUFS, but Spotify leaves 1 dB of headroom, so it can only raise it to about -16 LUFS before the peak ceiling stops it. No platform restores dynamics already crushed by a limiter, which is why mastering near the target beats mastering as loud as possible.

The loudness war, in brief

For two decades, roughly from the early 1990s through the 2000s, masters got steadily louder. The logic was simple and a little cynical: a louder track jumps out next to a quieter one on the radio or in a CD changer, so engineers pushed levels higher and squeezed dynamic range to win that perceived loudness. The cost was music that fatigues the ear and loses its punch, because nothing is quiet enough to make the loud moments hit.

Streaming normalization ended the incentive. When Spotify, Apple Music and YouTube play everything back to a common loudness target, a master made artificially loud is simply turned down to match, and all you are left with is the crushed dynamics. That is why the modern advice is to mix for dynamics and a sane target rather than for maximum loudness: the platform decides playback level, not your limiter.

Is -14 LUFS a hard rule? (the myth)

No. -14 LUFS is a playback reference, not a production target. Mastering exactly to -14 LUFS can leave a track flat, and pushing to -8 LUFS does not make it louder for the listener: the platform simply turns it back down, minus the punch already lost to compression. Master to the level the music needs, then check the integrated value and true peak before delivery.

Podcasts and audiobooks use different yardsticks

Spoken-word delivery does not follow the music streaming target. Podcasts aim for -16 LUFS in stereo and around -19 LUFS in mono, true peak -1 dBTP. Audiobooks do not use LUFS at all: the ACX standard behind Audible specifies RMS between -23 and -18 dB, peak no higher than -3 dB, and a noise floor below -60 dB RMS. If you are mastering narration, you are working to a separate spec entirely.

Broadcast and film: EBU R128

Television, radio and most film deliverables in Europe follow EBU R128, which fixes the program loudness rather than letting each platform normalize. North American broadcast follows the closely related ATSC A/85 standard.

• EBU R128: -23 LUFS integrated, true peak maximum -1 dBTP.
• ATSC A/85 (North America): -24 LKFS (LKFS and LUFS are numerically equivalent).
• Both also specify Loudness Range and a measurement gate, so always check the broadcaster's exact delivery sheet.

How to hit your target without crushing dynamics

The goal is to land on the right integrated loudness while keeping the dynamics the music needs. Use a loudness meter compliant with ITU-R BS.1770 on the master bus, read in integrated mode over the full track: iZotope Insight or RX, Youlean Loudness Meter, FabFilter Pro-L 2, or your DAW's built-in metering in Pro Tools.

In practice that means metering the whole program, setting a true-peak limiter at -1 dBTP, and using compression musically rather than as a loudness tool. When a master measures hotter than the reference, turning it down to the target by hand almost always sounds cleaner than letting the platform do it for you.

Delivery checklist for a streaming-ready master

Confirm the master against the targets before you send it.

• 24-bit WAV, 44.1 kHz or higher (88.2 / 96 kHz for Apple Digital Masters).
• Integrated loudness around -14 LUFS.
• True peak at -1 dBTP or lower (-2 dBTP if above -14 LUFS or for Amazon Music).
• No clipping, dynamics preserved.
• One master for all platforms, A/B checked against a reference track in the same genre.

Album vs track normalization on Spotify

Spotify normalizes loudness two ways depending on how you listen, and it changes how your tracks relate to each other. When you play a full album in order, Spotify normalizes the whole album as one unit, so the gain compensation stays the same across every track. The quiet interlude stays quiet and the loud single stays loud, exactly as the sequencing intended.

When tracks are played out of album context, such as in a playlist, on shuffle, or pulled from multiple albums, Spotify normalizes each track individually to the reference. That is why the same song can feel a touch louder or softer depending on whether it plays inside its album or inside a playlist. Master with both contexts in mind: a track that depends on album-relative loudness can sit differently once it lands in a playlist.

Spotify's three playback volume settings

On Spotify, Premium listeners can choose a normalization level, and each one targets a different reference loudness. This is the loudness your master is actually played back against, and it is why mastering far above the reference gains you nothing.

• Loud: around -11 LUFS, intended for noisy environments.
• Normal: around -14 LUFS, the default setting and the figure most mastering targets refer to.
• Quiet: around -19 LUFS, intended for quiet listening.

How to measure integrated LUFS

Measuring integrated loudness correctly takes a few minutes and removes the guesswork before delivery. The procedure below uses a meter compliant with ITU-R BS.1770.

• 1. Place an ITU-R BS.1770 compliant loudness meter as the last plugin on your master bus, after the limiter.
• 2. Set the meter to read integrated loudness (the value averaged over the whole program), not momentary or short-term.
• 3. Reset the meter, then play the track from the very start to the very end without stopping, so the integrated value covers the entire program.
• 4. Read the integrated LUFS figure and the true-peak maximum, and compare them to your target (around -14 LUFS, -1 dBTP for most streaming).
• 5. If you are over the target, lower the master gain to the target by hand rather than relying on platform normalization, then re-measure across the full track.
• Neutral meters that read ITU-R BS.1770 include Youlean Loudness Meter, FabFilter Pro-L 2 metering and many DAW built-in loudness meters. In our own workflow we measure with iZotope RX and inside Pro Tools.