What Sound Check does
Sound Check is Apple’s loudness normalisation: the loudness of each track is measured, and playback is adjusted so tracks come out at a consistent level. Like every service’s normaliser it is a playback gain change — the file you delivered is not altered. What Apple documents is attenuation: material measuring above the target is turned down toward it. Unlike Spotify, they do not publish a description of positive gain being applied to material below the target, so the safe assumption is that a quiet master plays at the level you delivered rather than being lifted to meet everything else.
Played as an album, the adjustment is applied to the album as a whole rather than per track, so the level relationships you mastered across a record survive; in a playlist, tracks are levelled individually. The general mechanics of loudness measurement — LUFS, gating, integrated versus short-term — are in What LUFS actually measures.
The practical consequence is the one worth internalising: on a normalised service, making a master louder does not make it play louder. It makes it play with less dynamic range at the same level. Every decibel of loudness you buy with compression is a decibel the normaliser gives back, and the compression stays.
Sound Check is a setting, not a guarantee
This is where Apple Music genuinely differs from Spotify, and most advice written about “mastering for streaming” flattens the difference away.
Spotify normalises by default. Sound Check is a user preference — a switch in the Music settings — and its default state has varied across operating-system versions and device types. You cannot know, for any given listener, whether it is on.
So Apple Music is the one major service where a hyper-loud master genuinely can play louder for part of the audience, because part of the audience has normalisation switched off. That is the honest version of the loudness-war argument, and it is a much weaker argument than it first appears:
- It only applies to listeners with it off, and you have no way to know or influence how many that is.
- It costs you on everyone else. Listeners with Sound Check on hear the same level and less dynamic range — a strictly worse version of your record.
- Playlists are where discovery happens, and a track that only holds up when it is louder than its neighbours does not hold up in a playlist at all.
- It does not transfer. The same master goes to Spotify and YouTube, which normalise regardless, so you are optimising one uncertain case at the cost of the certain ones.
Master the record to sound the way it should sound. That advice survives the setting being either way, which is the point of it.
The target, as published
- Apple Music — -16 LUFS integrated, ceiling -1 dBTP. Apple Digital Masters technical specification.
As on every platform, the target is where playback lands, not a number to master to. A −9 LUFS master and a -16 LUFS master both play at -16; the only difference the listener hears is how much dynamic range each has left. The figure comes from the document named above and platforms move these numbers occasionally, so check the source rather than trusting a number you learned once.
It is also worth saying what the target is not. It is not a minimum: a quiet, dynamic master sitting at −20 LUFS is not “wrong”, it simply gets turned up. It is not a maximum you will be rejected for exceeding — nothing refuses your upload for being loud. And it is not a mixing target. The number describes a finished master measured end to end, which is a different thing from where you set your faders. If those distinctions are new, mixing versus mastering is the place to start.
Albums are levelled as albums
Album normalisation is the feature most worth knowing about, because it is the one that protects deliberate artistic choices — and because it stops applying the moment a listener leaves the album.
When a record is played in sequence, one gain adjustment is calculated for the album and applied to every track. An interlude you mastered 6 LU below the single stays 6 LU below the single. A record that builds from a whisper to a wall still builds.
Pull one of those tracks into a playlist and it is levelled on its own. The quiet interlude is now turned up to sit beside everything else, which is usually not what it was for. Nothing is broken; the two contexts simply have different jobs, and it is worth deciding which one a given track is actually for before you master it.
What this means for sequencing
Master the album for the album, and check the singles as singles. If a track has to work in both places, the compromise is usually to keep its dynamics and accept that it plays quieter in the album context rather than to crush it so it survives a playlist.
What Apple Digital Masters actually is
“Apple Digital Masters” is not a loudness setting and not a file format you can select in a DAW. It is Apple’s certification program: mastering providers who meet Apple’s delivery specification — high-resolution sources, headroom for the encode, their validation tools — can be certified, and releases delivered through them carry the badge. The technical specification behind it is public, and it is the document the loudness and true-peak figures on this page come from.
Following the specification’s advice is available to anyone; the badge is not — it belongs to certified providers, which is why you will not find it used as a preset label here or anywhere honest. The advice itself is sound engineering regardless of certification: deliver from the highest-resolution master you have, and leave the encoder room to work.
The badge is also not an audible quality tier. A certified delivery and an ordinary one of the same master encode to the same thing. What the program actually buys is a checked chain — that nobody along the way handed the encoder a file already reduced to 16-bit 44.1 kHz and squashed to the ceiling.
True peak and the encode
Apple Music streams lossy encodes, and encoding can reconstruct inter-sample peaks higher than any sample in your file — which is why the ceiling is a true-peak figure (-1 dBTP) and why it is the strict number on this platform. A master that touches full scale on a sample-peak meter can clip inside the encode, after Sound Check has set the level, on the version of your track everyone actually hears.
Why a sample-peak meter misses it
A digital file stores samples, not a waveform. The waveform is what those samples reconstruct into on the way to a speaker, and between two samples that both sit at −0.2 dBFS the reconstructed curve can rise above 0. Your DAW’s meter shows the samples; the converter, and the encoder, produce the curve. A true-peak meter oversamples — rebuilding the signal at four times the rate or more — to see the peaks that are only there after reconstruction.
Lossy encoding makes it worse rather than better, because the encoder discards information and rebuilds an approximation whose peaks land in different places from the original. This is why the ceiling exists at all, and why −1 dBTP is a floor for caution rather than a generous allowance.
This is the reason to measure true peak specifically, with a meter that oversamples, rather than trusting the DAW’s peak reading. Our checker measures it to the same ITU-R BS.1770 standard the platforms use, in the browser, without uploading the file. If the master is already distorting rather than merely peaking, the clipping detector is the one that tells you where.
Spatial Audio is a separate delivery
Apple Music’s Spatial Audio catalogue is not the stereo master with an effect applied. It is a different deliverable — an Atmos master, authored from objects rather than channels — and it is measured and targeted separately from the stereo release.
Two consequences matter here. The loudness target for an Atmos delivery is not the stereo figure on this page, so do not carry the number across. And the stereo master is still what most listeners get, so it does not become a second-class citizen because an Atmos version exists.
What Dolby Atmos music actually is covers why object-based delivery behaves this way and why nearly everyone hears it as a binaural headphone render; our spatial audio page covers the render and export side.
−16 here, −14 there: does it matter?
Apple’s -16 and Spotify’s -14 look like a decision you have to make, and they are not. Each service measures your track and sets its own playback level independently — the same master lands on target on both, and neither is “quieter” in any way you can act on. There is no per-platform master worth making for loudness reasons alone; the Spotify page reaches the same conclusion from the other side.
The 2 LU gap does mean something, but not to you: it means Apple has chosen a slightly more conservative playback level than Spotify. Delivering a master 2 LU quieter to Apple would achieve nothing, because Apple would then turn it up by 2 LU. The hub page, streaming loudness targets, has the full table across services.
Mastering for Apple Music in practice
Master the track to sound right, keep true peak under the ceiling with a real true-peak limiter, and deliver from the highest-resolution source you have — the encoder gets more to work with from a 24-bit master than from something already reduced. Let Sound Check set the level; that is its job, and it does it the same way for everyone.
- Deliver 24-bit at the session rate. Do not pre-convert to 16-bit 44.1 kHz. Let the encoder work from the most information you have.
- Ceiling at -1 dBTP or below, measured with an oversampling meter, not the DAW’s sample-peak reading.
- Measure the finished file, not the mix bus in the session. Integrated loudness is a whole-programme measurement and the render is what ships.
- Check the album and the singles separately if the record is being released both ways.
- Do not master to −16. Master to what the song needs, then measure. The number is a readout, not a destination. How to master a song covers the order the decisions actually go in.
Euphona’s mastering ceilings true peak at −1 dBTP by default, measured with oversampling, and reports the measured loudness and true peak of every master — the values the platform will see, not the ones the session intended.
The questions people actually ask
What LUFS should I master to for Apple Music?
None in particular. Master so the record sounds right and keep true peak under -1 dBTP. If the result measures anywhere between roughly −14 and −8 LUFS you are in the range most commercial music occupies, and Sound Check will place it at -16 either way.
Does Apple Music turn quiet tracks up?
Apple documents Sound Check attenuating material above the target; nothing they publish describes it applying positive gain to material below it. Spotify does document raising quiet tracks, which is the safer thing to rely on if it matters to you. Either way there is no penalty for delivering a dynamic master — the worst case is that it plays at the level you delivered rather than being lifted.
Will a louder master beat a quieter one on Apple Music?
Only for listeners who have Sound Check off, and only at the cost of dynamic range for everyone who has it on. See the same question answered for Spotify, where the answer is a flat no because normalisation is not optional there.
Do I need a separate master for Apple Music?
No — not for loudness. One master serves every stereo streaming service. A separate deliverable is only required for the Atmos catalogue, which is a genuinely different format rather than a different level.
Why does my track sound quieter than a reference on Apple Music?
Almost always because the reference has more dynamic range rather than less. After normalisation both play at the same integrated loudness, so what you are hearing as “louder” is usually a stronger transient and a more open mix — a mix problem rather than a mastering one. Genre matters here too; mastering hip-hop works through a case where the low end is doing most of the work.
Keep reading