What Spotify does to your track
Spotify measures the integrated loudness of every track once, at ingest, and adjusts the playback level so tracks come out at a consistent loudness. The adjustment is a gain change applied while the track plays: your file is not re-encoded, re-mastered or altered, and switching normalization off plays the same file untouched.
The consequence is the one this whole subject reduces to: a master that measures louder than the target is turned down by exactly the difference. The loudness you paid for in dynamic range is removed at the door, and the dynamic range does not come back. The general mechanics (how the measurement works, what gating is, why platforms do this at all) are in What LUFS actually measures; this page is about what Spotify specifically does with the number.
One detail that surprises people: the measurement happens once, at ingest, and it is a property of the file you delivered. Re-delivering a different master means a new measurement. You cannot adjust how much gain Spotify applies except by changing the loudness of the master itself, and doing that changes nothing about the level anyone hears.
The target, as published
- Spotify, -14 LUFS integrated, ceiling -1 dBTP (-2 dBTP for masters louder than -14 LUFS). Published by Spotify in its guide for artists, “Loudness normalization”.
That figure is the default playback level, not an instruction to master to. A track at −9 LUFS and a track at -14 LUFS both play at -14; the difference between them after normalization is not loudness but how much dynamic range each still has. A track quieter than -14 is a different case: Spotify turns it up only until its true peak reaches -1 dBTP, so it may stay below the target. Platforms move these numbers occasionally, so the source named above is worth checking rather than trusting a figure you learned once.
Integrated loudness is measured across the whole track with quiet passages gated out, so it is not the average of the meter and it is not the loudest moment. A sparse verse does not drag the number down as far as you would expect, and a single loud chorus does not drag it up as far either. That is the measurement doing its job: it is trying to predict perceived loudness over the whole listen, which is the thing a normalizer needs to know.
Quiet, normal and loud
Normalization on Spotify is a listener setting with three levels, which Premium listeners can choose between. The same documentation that publishes the target puts them at −19 LUFS for quiet, −14 for normal and −11 for loud. Quiet and normal only ever turn tracks down or up by plain gain. The loud setting is the interesting one: a track measuring quieter than −11 has to be turned up to reach it, and positive gain can push peaks past full scale, so Spotify documents that it applies a limiter in that case to prevent clipping.
That limiter is the one scenario where Spotify processes your audio beyond a gain change, and it is a limiter you did not set, with a character you did not choose, applied to a master you cannot hear it on. It only engages on material quieter than −11 LUFS played on the loud setting, and it is a better argument than the loudness war ever was for delivering a master that does not need rescuing in either direction.
Which setting are your listeners on?
Normal is the default, and a default is what most of any audience keeps. The loud option exists mostly for noisy environments. It is not worth mastering for the edge cases, but it is worth knowing the loud setting is the one place a very dynamic master can have processing applied to it that you did not choose.
Albums are leveled as albums
Play a record in sequence and one gain adjustment is calculated for the album and applied across it, so the level relationships you mastered between tracks survive. The interlude you set 6 LU below the single stays 6 LU below the single.
Pull that interlude into a playlist and it is measured and leveled on its own, which turns it up toward everything else, as far as its peaks allow. Nothing is broken, the two contexts have different jobs. It is simply worth knowing which context a track is really for before deciding how dynamic to leave it. Apple Music behaves the same way, for the same reason.
When normalization is off
Listeners can switch normalization off, and then every track plays at whatever loudness it was mastered to. Some never get it at all: Spotify says its web player and third-party devices, like TVs and speakers, do not use loudness normalization. This is the only context in which a louder master is genuinely louder, and it cuts both ways, because it is also the context where an aggressive master sits next to a dynamic one and sounds small the moment the dynamic one’s chorus lands. The hub article covers how to reason about the on-and-off audience at once; the short version is that mastering for normalized playback does not harm playback without it, and the reverse is not true.
True peak on Spotify
Spotify streams lossy encodes, and a lossy encoder can reconstruct peaks slightly higher than the samples it was given, which is why the ceiling is specified as true peak (-1 dBTP) rather than sample peak. A master that clears the ceiling as measured by a sample-peak meter can still clip in the encode, where nobody hears it until it is public.
Why sample peak is the wrong meter
A 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 read −0.2 dBFS the reconstructed curve can rise above zero. Your DAW shows you 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 peaks that only exist after reconstruction.
The same Spotify document adds a stricter recommendation: masters louder than -14 LUFS should stay below -2 dBTP, because, in Spotify’s words, louder tracks are “more susceptible to extra distortion when encoded for streaming”. A true-peak reading is part of what our checker measures, on the same ITU-R BS.1770 standard the platform uses. If what you have is not headroom trouble but audible distortion already printed into the file, the clipping detector finds where it starts.
Three things that are not true
- “Spotify penalises loud masters.” There is no penalty and no ranking effect. A loud master is turned down by the amount it is loud. That is arithmetic, not punishment, and it applies identically to everyone.
- “You must master to −14 LUFS.” Spotify suggests it in its mastering tips, but nothing is rejected or flagged for missing it. A louder master is turned down to the same level as everything else.
- “Spotify re-encodes or re-masters your file.” It encodes for delivery, as every streaming service does, and applies playback gain. It does not apply EQ, compression or any tonal processing, with the single documented exception of the limiter on the loud setting described above.
Podcasts: no published figure
Spotify’s loudness guide talks only about music tracks. Its help pages for podcasts give no loudness figure, and Spotify does not say whether, or how, it normalizes podcast episodes. So do not carry the music number across. Dialogue also has a very different loudness distribution from music (long gaps, a narrow dynamic range, and gating that behaves differently) so the music figure is the wrong answer for speech anyway.
The practical advice for spoken word is mostly upstream of loudness anyway: a consistent level across the episode matters more than the integrated figure, and the fixes are restoration fixes. What we do for podcasters covers that path.
Mastering for Spotify in practice
Do not master to hit -14 LUFS exactly. The target is where playback lands, not a quality bar, and a master pushed or held to a number the mix did not ask for trades something real for a figure the normalizer makes irrelevant. Master the track to sound right, keep true peak under the ceiling, and let the platform set the level: one master serves Spotify, Apple Music and the rest, because each service adjusts it independently.
- Measure the rendered file, not the mix bus. Integrated loudness is a whole-program number and the bounce is what ships.
- Ceiling at -1 dBTP, or -2 dBTP if the master lands louder than -14 LUFS.
- Use an oversampling true-peak meter. The DAW’s sample-peak reading will tell you that you are fine when you are not.
- Fix loudness problems in the mix where they are actually caused. If the master needs heavy limiting to feel loud, the arrangement or the balance is what is fighting you: mixing versus mastering sets out which stage owns which problem.
- Deliver one master. Per-platform loudness masters solve a problem that normalization already solved.
Euphona’s mastering works the same way: the loudness target is chosen for the material and the destination, the published platform figures are the ones used, and the measured before-and-after loudness and true peak are in every job’s report, so what you deliver is a number you have seen, not one you hope for.
The questions people actually ask
What LUFS should I export for Spotify?
Whatever the finished master measures, provided true peak stays under -1 dBTP, or -2 dBTP if the master is louder than -14 LUFS. If you want a number rather than a non-answer, Spotify’s own mastering tips suggest -14 LUFS integrated. Louder masters are turned down to that level on the normal setting.
Is −14 LUFS too quiet?
No, and it is also not a level you need to hit. A master sitting at −14 gets no gain change on the normal setting; a master at −8 gets turned down 6 dB to the same place. The listener hears the same loudness from both.
Why does my song sound quieter than a commercial track?
If it measures louder than -14 LUFS, it is not quieter after normalization: it is less dense. What reads as “loud” at a matched level is transient impact, low-end weight and a mix with space in it, not integrated loudness. If it measures quieter, check its true peak first: Spotify lifts it only until the peak reaches -1 dBTP, so it can really play quieter. That is a mix and arrangement question; how to master a song covers what mastering can and cannot add, and mastering hip-hop works through a genre where the low end carries most of it.
Does Spotify use LUFS or ReplayGain?
The published target is stated in LUFS, measured to the ITU-R BS.1770 family of standards: the same measurement the broadcast world uses and the same one our checker implements.
Is the target different on Apple Music?
Apple publishes no loudness target for music. The -16 LUFS people quote is measured by third parties, and it changes nothing you would do. The Apple Music page explains why, and the one real difference: Apple says Sound Check can turn quiet tracks up, but not how far.
Keep reading