Streaming loudness targets for every platform (2026)

How streaming loudness normalization works

Every major streaming service measures a track's integrated loudness in LUFS (Loudness Units relative to Full Scale, defined by the ITU-R BS.1770-4 standard) and adjusts playback so songs sit at a consistent level. If your master is louder than the platform's reference, it gets turned down. Some platforms also turn a quiet master up — but only within limits, leaving headroom so the boost doesn't push true peaks into distortion after lossy encoding.

There are three LUFS measurements worth knowing:

• Integrated LUFS — the average loudness across the whole track. This is the number platforms normalize to, and the number you master to.
• Short-term LUFS — a ~3-second rolling average, useful for watching dynamics section to section.
• Momentary LUFS — a ~400 ms window, useful for catching transient-heavy moments.

The practical consequence: mastering louder than the reference buys you nothing on most platforms. A −8 LUFS master and a −14 LUFS master are turned down to the same integrated playback level on Spotify — but the −8 master sacrificed its dynamics (lower crest factor, audible limiting artifacts) to get there, and ends up sounding flatter, not louder. The loudness war is over, and normalization won it.

Platform loudness targets at a glance (2026)

Platform	Reference loudness	True-peak ceiling	Adjusts
Spotify	−14 LUFS	−1 dBTP (−2 if > −14 LUFS)	Up & down
Apple Music	−16 LUFS	−1 dBTP	Up & down (Sound Check)
YouTube / YT Music	−14 LUFS	−1 dBTP	Down only
Tidal	−14 LUFS	−1 dBTP	Up & down
Amazon Music	−14 LUFS	−2 dBTP	Up & down
Deezer	−15 LUFS	−1 dBTP	Up & down
TikTok	~−14 LUFS (rolling out)	−1 dBTP	Increasingly down
Bandcamp	No normalization	—	Plays as-is

Read these as a center-of-range, not law. References shift over time and vary by device and user setting. Spotify alone offers three modes: Loud (~−11 LUFS, applies a limiter), Normal (−14 LUFS), and Quiet (~−19 LUFS). Apple's Sound Check targets a ~−16 LUFS baseline and adjusts in both directions, but can behave inconsistently on Lossless/Dolby Atmos tracks and across devices. TikTok historically did not normalize in-feed; that is changing in 2026, but enforcement is uneven, so a louder master can still grab attention scrolling past.

So what should I master to?

For most modern releases, aim for an integrated loudness around −14 LUFS with true peak at or below −1 dBTP. That keeps you at or just under the level Spotify, YouTube, Tidal and Amazon normalize to — nothing gets turned down — while leaving codec headroom. (For Amazon specifically, keep true peak under −2 dBTP.)

Target by genre

Genre / context	Suggested integrated LUFS	Why
Classical, jazz, acoustic, ambient, orchestral	−16 to −18 LUFS	Dynamic range is the point; YouTube won't turn it up, and the dynamics survive normalization everywhere else
Indie, folk, singer-songwriter, trip-hop, downtempo	−13 to −15 LUFS	Preserves breath and groove while staying near reference
Pop, rock, electronic	−11 to −14 LUFS	Genre feel; anything above −14 is turned down on most platforms
Hip-hop, trap, EDM, metal	−9 to −12 LUFS	Density is part of the sound — master for feel, accept the normalization penalty

Album consistency matters more than any single number. Master the record so tracks feel coherent relative to one another, then check the loudest track against the platform reference.

Rule of thumb: master for the song and the genre, keep true peak under −1 dBTP (−2 for Amazon), and stop chasing loudness past −14 LUFS — the platforms give it all back.

What the turn-down actually does

It's easy to nod along with "the platforms turn you down" and still master hot. So make it concrete. Here's a −7.5 LUFS master — squarely in loudness-war territory — and where each platform actually lands it:

Platform	Adjustment	Plays your master at
Spotify (Normal)	−6.5 dB	−14 LUFS
Apple Music	−8.5 dB	−16 LUFS
YouTube	−6.5 dB	−14 LUFS
Tidal	−6.5 dB	−14 LUFS
Amazon Music	−6.5 dB	−14 LUFS
Deezer	−7.5 dB	−15 LUFS
Bandcamp	none	−7.5 LUFS

The only place that master is actually louder than the room is Bandcamp, which doesn't normalize. Everywhere people actually stream, a clean −14 LUFS master would have landed at the identical playback level — with its transients, depth and crest factor intact, and without the inter-sample clipping a hot true peak invites once a codec gets hold of it. You can hear this exact turn-down on your own file with the loudness tool before you commit to it.

Why true peak, not sample peak

A limiter set to 0 dBFS at the sample level still produces inter-sample peaks — overshoots that appear when the signal is reconstructed by a converter or re-encoded by a lossy codec (AAC, Ogg Vorbis). A file that reads 0.0 dBFS on a sample meter can measure +1 dBTP or more after Spotify or Apple encodes it, which clips and adds crackle on transients — drum hits, vocal consonants, synth stabs.

That's why every platform specifies a true-peak ceiling, measured with oversampling per ITU-R BS.1770, and why leaving −1 dBTP (−2 for Amazon, or for any master hotter than −14 LUFS on Spotify) is standard practice. See LUFS, true peak & dynamic range explained for the full picture.

Common mistakes that quietly hurt your master

• Mastering to the "Loud" mode number (−11). Most listeners are on Normal. Targeting −11 just sacrifices dynamics for a setting most people never use.
• Reading sample peak instead of true peak. 0.0 dBFS is not safe. Always meter dBTP with oversampling.
• Ignoring the Amazon −2 dBTP exception. A master that's clean everywhere else can distort on Amazon if it sits at −1 dBTP.
• Chasing a single number across an album. Relative balance between tracks beats hitting −14.000 on every one.
• Assuming TikTok rewards maximum loudness forever. It's normalizing more in 2026 — over-limiting for the feed will increasingly just cost you dynamics.

How to check your master in seconds

You don't have to guess. Drop a WAV or FLAC into the free Soneam loudness tool and it shows your integrated LUFS, true peak and dynamic range — and plays your master back the way each platform will after normalization, so you can hear the turn-down before you deliver. Nothing is uploaded or stored; analysis runs in your browser.