LUFS targets per platform in 2026: Spotify, Apple Music, YouTube, Netflix, broadcast

Published: 2026-06-05 · Reading time: 23 min read · Author: Nikolay Sapunov, CEO at Fora Soft

Why this matters

If you ship audio anywhere — a music release, a podcast, a YouTube channel, an OTT catalogue, a broadcast feed — someone hands you a number and expects you to hit it, and the number is different on every platform. Master a podcast to −23 LUFS and listeners reach for the volume; master a single to −7 LUFS to "sound loud on Spotify" and the platform turns it down anyway, so you traded away dynamic range for nothing. This article is for the product manager, founder, content lead, or operations person who needs to know the right target for each destination, understand why broadcast and streaming live in different worlds, and be able to read or write a delivery spec without guessing. Every number here traces to a controlling document — the platform's own published spec, an ITU-R Recommendation, an EBU R-document, an ATSC Recommended Practice, or the AES streaming recommendation — not to a forum post.

What a "LUFS target" actually is

Before the table, one idea has to be solid, because everything else hangs off it. The loudness number, called LUFS — Loudness Units relative to Full Scale — is a measurement of how loud a piece of content sounds to a human ear, averaged over its whole length, expressed as a negative number where closer to zero means louder. A track at −14 LUFS is louder than a track at −23 LUFS. The measurement comes from an international algorithm, ITU-R BS.1770, that filters the audio to match human hearing, ignores the silent passages, and boils the whole thing down to one figure. If that sentence is new to you, read the measurement piece first: Loudness Normalization: EBU R128, ITU-R BS.1770, ATSC A/85. This article assumes you know what a LUFS number is and focuses on the targets — the specific numbers each platform aims for.

A "target" is the loudness a platform wants every item in its catalogue to play back at. Here is the part that trips people up: the platform enforces the target at playback, not by rewriting your file. When you upload a song or video, the platform measures its integrated loudness — the gated, whole-programme LUFS figure — and stores it. At play time it applies a single, uniform gain change to push that item toward the target. If your master measures −9 LUFS and the target is −14 LUFS, the platform plays it 5 dB quieter. Your file on disk is untouched; only the playback level moves.

Because it is one uniform gain move, normalization never adds distortion and never changes the relationships inside your mix — the quiet parts and loud parts keep their exact distance from each other. It just slides the whole thing up or down to land on the target. This is the single fact that defeats the "master louder to win" instinct: you cannot out-loud the normalizer, because it measures you and cancels the excess.

The −14 vs −23 confusion, killed early. The most repeated myth in audio is that there is one loudness standard. There is not. −23 LUFS is the European broadcast target. −14 LUFS is the typical streaming target. −27 LKFS is Netflix's dialog-gated cinematic target. They are all correct — for different destinations. The question is never "what is the LUFS standard?" It is always "what is the target for this platform?"

The one table: 2026 LUFS targets per platform

This is the table the rest of the article exists to justify. Every row is sourced; the platform's own published value wins over any third-party estimate. Where a platform has never published an official number, the row says so explicitly — guessing in a delivery spec is how projects fail.

Table 1. Consolidated 2026 LUFS targets. Two different jobs share one table: the streaming/music rows are normalization targets enforced at playback; the broadcast and Netflix rows are delivery specs you must hit before handing over a file. LKFS and LUFS are the same unit with two names (US vs Europe). Sources are listed in full under References; "no published target" means exactly that — do not write a fixed number into a contract for those platforms.

Two things to notice immediately. First, almost every music platform converges on −14 LUFS, with Apple the outlier at −16 — that convergence is recent and deliberate, driven by the AES recommendation the platforms helped write. Second, the broadcast and Netflix rows are 9 to 13 LU quieter than streaming, which is not a rounding error; it is a different philosophy that the rest of the article explains.

Figure 1. The 2026 LUFS landscape on one axis. Three clusters: broadcast/cinema (quiet, wide dynamics), streaming/podcast (the −14 to −18 mainstream), and the unofficial social-media "hot zone". Louder is to the right.

Why streaming is louder than broadcast

The numbers split into two camps — broadcast and cinema near −23 to −27, streaming near −14 to −16 — and the gap looks like a disagreement. It is not. Both camps measure with the same BS.1770 ruler and both want the same outcome: no surprises for the listener. They land on different numbers because they are solving for different rooms.

Broadcast and cinema assume an attentive listener in a quiet, controlled space — a living room at night, a cinema auditorium. In that room, quiet passages are audible and loud passages are welcome, so the content can have a wide dynamic range: the gap between the whisper and the explosion. A quieter overall target like −23 LUFS leaves headroom for those loud moments without forcing the whole mix to be dense. The quiet target is a feature: it protects dynamics.

Streaming assumes the opposite room. You are on a phone or earbuds, on a train, in a café, walking down a street with traffic. Background noise swallows quiet passages, and you do not want to ride the volume knob between songs. So streaming normalizes to a louder, denser target — −14 LUFS — where even the quiet parts sit above the noise floor of the world around you. The louder target is also a feature: it suits mobile listening.

Here is the arithmetic that makes the gap concrete. Take the same master and deliver it to two destinations. To go from a streaming target of −14 LUFS to a broadcast target of −23 LUFS, you change the level by:

gain = broadcast target − streaming target
gain = −23 LUFS − (−14 LUFS)
gain = −9 LU

Nine loudness units — roughly nine decibels — separates the two worlds. That is not a tweak; it is the difference between "comfortable on earbuds in a noisy street" and "comfortable in a silent cinema". The same content needs two different deliveries, and the only thing that changes is one uniform gain move plus a peak check.

Pitfall — assuming one master fits every destination. A −14 LUFS, −1 dBTP music master is a sensible single deliverable for the music-streaming platforms, because they share a target. But the same file is wrong for EBU broadcast (−23), wrong for Netflix (−27 dialog-gated, −2 dBTP), and over-compressed for a quiet-living-room cinema mix. "One master to rule them all" works within the streaming cluster and breaks the moment you cross into broadcast or cinema. Match the master to the destination's row in Table 1.

The music-streaming cluster: −14 LUFS and one outlier

The music services have quietly converged. Spotify, YouTube, TIDAL, Amazon Music, and SoundCloud all normalize to −14 LUFS integrated. Deezer sits one unit quieter at −15. Apple Music is the deliberate outlier at −16 LUFS. That convergence is not coincidence — the major services collaborated on the AES TD1008 recommendation (covered below), which nudged the industry toward a consistent zone so a listener moving between apps no longer hears a loudness jump.

The practical upshot for anyone delivering music: a single master at −14 LUFS integrated with a true peak no higher than −1 dBTP is the right deliverable for the −14 cluster, and it plays acceptably on Apple's quieter −16 target because Apple mostly turns content down rather than up. Mastering louder than −14 buys you nothing on these platforms — they measure your file and apply negative gain to bring it back to target, so the extra loudness is discarded and you keep only the reduced dynamic range you traded for it.

Spotify's three modes — a useful detail most specs omit

Spotify is worth a closer look because it exposes the normalization that other platforms hide. A Spotify Premium listener can choose one of three loudness modes in the app settings, and each maps to a different target:

Table 2. Spotify's three playback loudness modes. The default is −14 LUFS, which is why −14 is the number to master for. The "Loud" mode is the only common case where a platform applies a limiter on top of gain — which can audibly squash an otherwise dynamic master.

The "Loud" mode is the one detail that justifies keeping a little headroom in your master. In Loud mode Spotify must raise quiet content toward −11 LUFS, and to stop the loudest peaks from clipping it inserts a limiter. A master that was mixed with no headroom and pushed to the ceiling can be audibly squashed by that limiter. Leaving a true-peak ceiling of −1 dBTP (or −2 dBTP on upload, which Spotify recommends) gives the platform room to work without distortion.

Podcasts and spoken word: the −16 LUFS world and the mono trap

Spoken-word content lives at a slightly quieter target than music, and for a measurable reason. Apple Podcasts recommends −16 LUFS (±1 dB); Spotify's podcast guidance is −14 LUFS; the broad industry convention for podcasts is −16 LUFS. The reason speech targets sit a touch lower than the −14 music number is a quirk of the BS.1770 meter: it reads speech and music slightly differently, and listening tests show speech normalized to the same BS.1770 number as music is perceived as 2 to 3 dB louder. So the standards bodies recommend normalizing speech a few units quieter than music to make them feel equally loud.

There is a mono trap here that bites podcasters constantly. A LUFS measurement of a mono file and a stereo file are not directly comparable, because the meter sums channels. To sound as loud as a stereo file at −16 LUFS, a mono file should be measured at roughly −19 LUFS. Deliver a mono podcast at −16 LUFS and it will play back noticeably louder than the stereo episodes around it.

mono target that matches a −16 LUFS stereo file ≈ −16 LUFS − 3 LU = −19 LUFS

Most talk podcasts are mono (smaller files, identical on every device), so the right number for a mono interview show is −19 LUFS, not −16. Use stereo and −16 LUFS only when music and sound effects justify the width. Getting this wrong is the single most common loudness complaint in podcast feeds.

Broadcast and cinema: the quiet, regulated end

The professional video and broadcast world is where loudness targets carry the weight of law and contract, and the numbers are far quieter than streaming.

EBU R128 — the European broadcast rulebook — sets the target at −23 LUFS, with a tolerance of ±0.5 LU for file-based delivery and a −1 dBTP true-peak ceiling. ATSC A/85 — the US equivalent — sets −24 LKFS and is enforced by federal law through the CALM Act, which makes the target legally binding for television commercials. The one-unit gap between −23 and −24 is historical, not perceptual; a single mix delivered to both needs only a one-decibel trim. Both rest on the same BS.1770 measurement; the full mechanics are in Loudness Normalization: EBU R128, ITU-R BS.1770, ATSC A/85.

Netflix is the most-cited streaming-video example, and it does something distinctive: its delivery spec asks for −27 LKFS, dialog-gated, with a ±2 LU tolerance and a −2 dBTP true-peak ceiling, measured per BS.1770. "Dialog-gated" means the loudness is measured only during sections where dialogue is present — an action sequence with no talking does not pull the number down. If a programme lacks enough dialogue for the gating to work, Netflix falls back to a full-programme integrated measurement near −24 LKFS. The −27 dialog-gated figure looks dramatically quieter than streaming music's −14, but it is measuring a different thing (dialogue only, not the whole mix) and serving a different room (a cinematic mix with wide dynamics on a good system). Netflix's approach is documented in its public partner specifications, and the dialog-anchored philosophy mirrors the ATSC A/85 "Anchor Element" idea.

Pitfall — comparing a dialog-gated number to a whole-programme number. Netflix's −27 LKFS and Spotify's −14 LUFS are not on the same scale of comparison even though both use BS.1770. Netflix gates to dialogue; Spotify measures the whole track. A film whose dialogue sits at −27 LKFS can have an overall integrated loudness several LU louder once the music and effects are counted. Never line up a dialog-gated target against a whole-programme target as if they measured the same thing.

The AES recommendation that pulled streaming together

The convergence around −14 to −16 LUFS did not happen by accident. The Audio Engineering Society published a technical recommendation — AES TD1008 (document AESTD1008.1.21-9, "Recommendations for Loudness of Internet Audio Streaming and On-Demand Distribution", 2021, superseding TD1004 from 2015) — and the major services helped draft it. It is the closest thing streaming has to a shared rulebook, and it sets content-aware targets rather than one universal number:

Table 3. AES TD1008 content-aware streaming recommendations. The document deliberately recommends normalizing speech 2–3 LU quieter than music so the two feel equally loud through the BS.1770 meter. Source: AES TD1008 (AESTD1008.1.21-9, 2021).

TD1008 also recommends an album-normalization method: normalize the loudest track of an album to −14 LUFS, which typically yields an album-integrated loudness near −16 LUFS and keeps most quieter tracks above −20 LUFS. That album-aware logic is exactly what Spotify and TIDAL implement when they preserve the volume relationships between tracks on the same record instead of flattening every track to an identical number.

Social media: the platforms that won't tell you

TikTok, Instagram, Facebook, X, and Twitch are the honest gap in any LUFS table, and a responsible spec says so. None of them publishes an official LUFS target. Every fixed number you see online for these platforms is an estimate. The best-supported public information is that Meta moved to xHE-AAC with adaptive loudness management — the platform adjusts loudness to the playback context (headphones versus speaker, quiet versus noisy) rather than normalizing to one fixed figure. Observed behaviour clusters internal targets somewhere around −14 LUFS, but it is not a number you can hold the platform to.

The practical consequence is a real tension. Because enforcement on these platforms is loose and inconsistent, many creators master hot — −10 to −12 LUFS — to grab attention in the first three seconds of a feed scroll on a phone speaker, accepting whatever normalization penalty applies. That is a content strategy, not a standard. If you need a defensible number for a social deliverable, target around −14 LUFS with peaks held at −1 dBTP, and treat anything louder as a deliberate, eyes-open trade rather than spec compliance.

What happens when you break a target

It helps to know the failure mode for each direction, because they are different.

Too loud is the common case. You master at −8 LUFS to "compete". The platform measures −8, the target is −14, so it applies −6 dB of gain at playback. Your track now plays at −14 LUFS like everyone else's — but you spent dynamic range to get loud, and that range is gone forever; the normalizer cannot give it back. You sound the same loudness as a track mastered at −14, but more squashed. Net result: you lost and gained nothing.

playback gain = target − your measured loudness
playback gain = −14 LUFS − (−8 LUFS)
playback gain = −6 dB  (turned down, dynamics already sacrificed)

Too quiet behaves differently depending on the platform. Spotify, TIDAL, Amazon, and Deezer will raise a quiet master toward the target. YouTube and Apple's Sound Check largely only turn down — a master quieter than target simply plays quiet, with no boost. So a −20 LUFS master plays at −14 on Spotify but stays at −20 on YouTube, sounding weaker than the videos around it. If your destinations include turn-down-only platforms, do not deliver well below target expecting a free boost.

Peaks over the ceiling is the third failure. A master that peaks at 0 dBFS may show no clipping in your editor, but the true inter-sample peak — the real analog level between digital samples — can sit above 0 and clip on consumer hardware, especially after the platform's encoder reconstructs the waveform. That is why every row in Table 1 carries a true-peak ceiling (−1 dBTP for streaming, −2 dBTP for broadcast and Netflix). We go deep on this in True Peak, dBTP and the Inter-Sample Peak Problem.

Figure 2. What the normalizer does to an off-target master. Too loud: turned down, dynamics already lost. Too quiet: boosted on some platforms, left weak on turn-down-only ones.

A decision tree: which target do I use?

Pull it together into a routine you can run for any deliverable. Identify the destination, read the target, set the peak ceiling, and decide whether one master covers more than one destination.

Figure 3. Pick the destination, read the target and ceiling. Most music destinations share −14 LUFS / −1 dBTP; everything else is its own row.

The routine in words: for music streaming, master once at −14 LUFS / −1 dBTP and it covers Spotify, YouTube, TIDAL, Amazon, SoundCloud, and plays fine on Apple's −16. For a podcast, target −16 LUFS stereo or −19 LUFS mono, −1 dBTP. For European broadcast, −23 LUFS / −1 dBTP; for US broadcast, −24 LKFS. For Netflix, follow the −27 LKFS dialog-gated, −2 dBTP delivery spec exactly. For social media, treat −14 LUFS as a sensible default and accept that the platform may do something adaptive you cannot fully predict.

Where Fora Soft fits in

Across the OTT, video streaming, e-learning, telemedicine, and video conferencing products we have built since 2005, the per-platform loudness question shows up as an integration detail that quietly decides whether a service feels professional. For OTT and streaming clients we build the BS.1770 measurement and a true-peak limiter into the transcode step, then route each rendition to the correct target — streaming, broadcast, or a partner's delivery spec — so no asset reaches the player off-target. For e-learning and telemedicine, where a quiet lecturer and a loud one must arrive at the listener at the same comfortable level, we normalize captured audio to a single consistent target before delivery. The standards do not change between projects; what changes is which row of the table the market demands, and wiring the right target into the pipeline rather than checking it by hand at the end is the difference between a catalogue that behaves and one that makes listeners reach for the volume.

What to read next

• Loudness Normalization: EBU R128, ITU-R BS.1770, ATSC A/85
• Loudness, Peak, RMS, LUFS — Measuring How Loud Audio Actually Is
• True Peak, dBTP and the Inter-Sample Peak Problem

Call to action

• Talk to a audio engineer — book a 30-minute scoping call to talk through your lufs targets per platform plan.
• See our case studies — 250+ shipped projects across video streaming, WebRTC, OTT, telemedicine, e-learning, surveillance, and AR/VR.
• Download the 2026 LUFS targets per platform — cheat sheet — One-page reference: the integrated LUFS target and true-peak ceiling for every major destination (Spotify, Apple Music, YouTube, TIDAL, Amazon, Deezer, podcasts, Netflix, EBU, ATSC), Spotify's three modes, the mono-vs-stereo podcast….

References

1. Spotify, "Loudness normalization on Spotify" (Spotify for Artists Help, accessed 2026-06-05). First-party controlling source for Spotify's −14 LUFS Normal target, the Loud (−11 LUFS) / Normal (−14) / Quiet (−19) playback modes, the limiter applied in Loud mode, album-aware normalization, and the −1 / −2 dBTP upload guidance. https://support.spotify.com/us/artists/article/loudness-normalization/ — Tier 4 (first-party platform documentation; the controlling source for Spotify's own target).
2. Apple, "Apple Music and Sound Check loudness" (production-expert.com analysis of Apple's −16 LUFS choice and its AES TD1008 alignment; Apple Sound Check enabled by default since 2022, accessed 2026-06-05). Source for Apple Music's −16 LUFS reference level and the turn-down-mostly behaviour. https://www.production-expert.com/production-expert-1/apple-choose-16lufs-loudness-level-for-apple-music-heres-why — Tier 4 (deployer documentation / first-party-derived; Apple's published value).
3. YouTube loudness normalization (−14 LUFS, turn-down-only) (Critical Listening Lab platform reference, cross-checked against YouTube "Stats for nerds" volume/normalized readout, accessed 2026-06-05). Source for YouTube's −14 LUFS target and its turn-down-only behaviour. https://www.criticallisteninglab.com/en/learn/loudness/youtube — Tier 4 (deployer-behaviour documentation; the observable YouTube target).
4. Netflix, "Loudness and True Peaks: How to Measure and When to Flag" and "Netflix Sound Mix Specifications & Best Practices v1.6" (Netflix Partner Help Center, accessed 2026-06-05). First-party controlling source for Netflix's −27 LKFS dialog-gated target, the ±2 LU tolerance, the −2 dBTP true-peak ceiling, the BS.1770 measurement basis, and the −24 LKFS full-programme fallback when dialogue is insufficient for gating. https://partnerhelp.netflixstudios.com/hc/en-us/articles/360050414014 — Tier 4 (first-party deployer delivery specification).
5. AES, Technical Document AESTD1008.1.21-9, "Recommendations for Loudness of Internet Audio Streaming and On-Demand Distribution" (Audio Engineering Society, 2021; supersedes TD1004:2015; accessed 2026-06-05). Controlling source for the content-aware streaming recommendations: −16 LUFS music, −18 LUFS news/talk/dramatic, −17 LUFS mixed/sports, the album-normalization method (loudest track to −14 LUFS), and the rationale that speech is perceived 2–3 LU louder than music at the same BS.1770 reading. https://www.aes.org/technical/documentDownloads.cfm?docID=731 — Tier 5 (AES technical recommendation; the cross-platform reference the major services helped draft). The PDF is served as a non-fetchable download in our pipeline; figures cross-confirmed against the AES catalogue entry and multiple independent summaries.
6. EBU, R 128, "Loudness normalisation and permitted maximum level of audio signals", v5.0 (European Broadcasting Union, November 2023, accessed 2026-06-05). Controlling source for the −23 LUFS European broadcast target, the ±0.5 LU tolerance, the −1 dBTP ceiling, and the R128 s2 "Loudness in Streaming" supplement explaining louder mobile targets. https://tech.ebu.ch/publications/r128 — Tier 1 (official EBU Recommendation).
7. ITU-R, Recommendation BS.1770-5, "Algorithms to measure audio programme loudness and true-peak audio level" (ITU-R, November 2023, accessed 2026-06-05). Controlling source for the LUFS/LKFS measurement algorithm (K-weighting, gating, integration) and the dBTP true-peak definition that underpins every target in this article. https://www.itu.int/rec/R-REC-BS.1770 — Tier 1 (official ITU-R Recommendation; the measurement engine all targets share). The ITU PDF is a direct download not fetchable in our pipeline; values are cross-confirmed via the ITU catalogue page and the EBU/ATSC documents that cite it.
8. ATSC, A/85:2013 (with Corrigendum No. 1, February 2021), "Techniques for Establishing and Maintaining Audio Loudness for Digital Television" (Advanced Television Systems Committee, accessed 2026-06-05). Controlling source for the −24 LKFS US broadcast target, the Anchor Element (dialogue) approach that parallels Netflix's dialog gating, and the −2 dBTP recommended ceiling; made mandatory for ads by the CALM Act. https://www.atsc.org/atsc-documents/a85-techniques-for-establishing-and-maintaining-audio-loudness-for-digital-television/ — Tier 1 (official ATSC Recommended Practice).
9. TIDAL, Amazon Music, SoundCloud, Deezer loudness behaviour (iZotope, "How to master for streaming platforms: normalization, LUFS, and loudness", accessed 2026-06-05). Source for the −14 LUFS targets of TIDAL, Amazon Music, and SoundCloud, and Deezer's −15 LUFS, plus album-aware behaviour. https://www.izotope.com/en/learn/mastering-for-streaming-platforms — Tier 4 (vendor engineering reference compiling published platform values).
10. Apple Podcasts and podcast loudness convention (−16 LUFS / ±1; mono −19 LUFS) (Podnews, "How to do loudness: the LUFS and LKFS FAQ for podcasters"; Apple Podcasts for Creators audio requirements; accessed 2026-06-05). Source for the −16 LUFS podcast recommendation, the ±1 dB tolerance, and the mono-vs-stereo −19 LUFS equivalence. https://podnews.net/article/lufs-lkfs-for-podcasters — Tier 4 (first-party Apple guidance plus deployer reference).
11. Meta / social-media adaptive loudness (no published target) (independent platform analysis confirming Meta's xHE-AAC adaptive loudness and the absence of any official TikTok/Instagram/Meta LUFS figure, accessed 2026-06-05). Source for the explicit "no official target" status and the context-adaptive behaviour. https://danmurtagh.com/lufs-loudness-standards/ — Tier 6 (aggregator; used only to document the absence of an official spec, never as a source of a fixed number).