/ˈɪn.tɪ.ɡreɪ.tɪd ˈlaʊd.nəs/
Integrated Loudness is the time-averaged perceived loudness of a complete audio program, measured in LUFS (Loudness Units relative to Full Scale). It is the primary loudness metric used by streaming platforms to normalize playback volume across all content.
The loudness wars didn't end because producers got tired of squashing dynamics — they ended because a measurement standard finally gave mastering engineers something concrete to aim for, and platforms something real to enforce.
Integrated Loudness is the single-number representation of how loud a complete audio program sounds across its entire duration. Expressed in LUFS (Loudness Units relative to Full Scale) or equivalently in LU (Loudness Units) relative to a reference, it is computed by the algorithm defined in ITU-R BS.1770 and refined in EBU R128. Unlike peak meters — which capture instantaneous ceiling breaches — or RMS meters — which average power over short windows — Integrated Loudness measures perceived loudness from the first non-silent sample to the last, gating out passages that fall more than 10 LU below the ungated average. The result is a perceptually weighted, program-wide number that closely tracks what human ears experience as overall volume.
The practical significance for producers and mastering engineers is enormous. Every major streaming platform — Spotify, Apple Music, YouTube, Tidal, Amazon Music, Deezer — applies loudness normalization that references the integrated measurement of your uploaded file. Spotify targets −14 LUFS integrated; Apple Music and Tidal normalize to −16 LUFS; YouTube normalizes to −14 LUFS for music content; Amazon Music HD sits around −14 LUFS. When a track exceeds a platform's target, the platform turns it down. When a track is quieter, some platforms turn it up (Spotify with normalization on), and some leave it alone (Spotify with normalization off). Either way, the arms race of over-limiting to sound louder than competitors collapses: if both tracks are normalized to −14 LUFS, the louder master gains nothing in perceived volume but loses everything in dynamic range.
Integrated Loudness differs in an important conceptual sense from its sibling metrics. Short-term loudness measures a 3-second sliding window and is useful for monitoring vocal levels and mix consistency in real time. Momentary loudness measures a 400 ms window and catches transient energy spikes. True Peak measures the maximum interpolated peak between digital samples, guarding against inter-sample clipping introduced during lossy encoding. Integrated Loudness is the only one of these that collapses an entire program into one number, making it the metric of record for delivery and normalization decisions. A mastering session that ignores this number is a mastering session flying blind.
The K-weighting filter built into the BS.1770 measurement algorithm is why integrated loudness correlates so well with perception. The filter applies a high-shelf boost above roughly 2 kHz — prioritizing mid and upper frequencies where the ear is most sensitive — and a low-frequency rolloff below about 60 Hz, where sub-bass energy has little perceived loudness impact relative to its power contribution. This means a mix loaded with 50 Hz sub content will measure quieter in LUFS than its RMS reading suggests, while a bright, harsh mix will measure louder. Producers who understand K-weighting use it diagnostically: if a master is hitting −14 LUFS integrated but still sounds thin or harsh, the spectral balance is the culprit, not the limiter.
Understanding integrated loudness does not mean simply targeting a number and calling it done. It means using the metric as a feedback loop: render a mix, measure integrated loudness, compare to the platform target, adjust gain staging or limiting, and re-check dynamics. A well-mastered track at −14 LUFS integrated with a true peak ceiling of −1.0 dBTP will sound competitive on every platform, preserve transient detail, and survive the re-encoding to AAC or OGG Vorbis without clipping artifacts. That is the practical promise of the integrated loudness standard — not a creative constraint, but a shared reference point that levels the playing field.
The measurement algorithm defined in ITU-R BS.1770 processes audio in overlapping 400 ms blocks advanced in 100 ms steps, effectively creating a stream of momentary loudness values throughout the program. Each block is passed through K-weighting — a two-stage filter consisting of a pre-filter that models the acoustic effect of the head (a high-shelf boost of approximately +4 dB centered around 1.5 kHz) and a revised low-frequency filter (a second-order high-pass at roughly 38 Hz). Mean square values are then computed per channel, scaled by a channel weighting factor (center and surround channels receive different weights in multichannel configurations; for stereo both channels are weighted equally at 1.0), and summed. The result is a momentary loudness value in LUFS for that 400 ms block.
Integrated loudness is derived from these momentary values through a two-stage absolute and relative gating process. The absolute gate discards any 400 ms block whose loudness falls below −70 LUFS — periods of silence or near-silence that would otherwise drag the average down artificially. The relative gate then computes a preliminary integrated loudness from all blocks that passed the absolute gate, subtracts 10 LU from that value, and discards any block falling below this new threshold. This second gate removes quiet passages — audience noise between movements, a sparse verse that dramatically undercuts the program average — ensuring the final integrated number reflects the loudness of the musically active content. Only after both gates are applied does the algorithm compute the true time-averaged mean loudness, expressed in LUFS.
True Peak measurement operates on a parallel pathway. The BS.1770 specification requires upsampling audio to at least four times the original sample rate before detecting peaks, because inter-sample peaks — reconstructed analog values between two digital samples — can exceed 0 dBFS after D/A conversion or lossy encoding even when every individual sample reads below 0 dBFS. The industry standard for delivered masters is a True Peak ceiling of −1.0 dBTP, though some engineers use −0.5 dBTP for lossless delivery and −1.5 dBTP for formats that will undergo heavy lossy re-encoding. A limiter that only constrains sample peaks without True Peak limiting is not sufficient for streaming delivery.
Modern loudness meters implement the full BS.1770-4 specification and display all four loudness metrics simultaneously: Momentary (M), Short-Term (S), Integrated (I), and Loudness Range (LRA). LRA, defined in EBU Tech 3342, measures the statistical spread of loudness over the program using the 10th and 95th percentile of the short-term distribution — it quantifies dynamic range in a perceptually relevant way. A highly compressed pop master might show LRA of 4–6 LU; an orchestral recording might show 14–20 LU. Monitoring all four values during mastering reveals the full dynamic picture that no single metric captures alone.
For producers working in a DAW, integrating loudness measurement into the signal chain means placing a compliant meter — one reading BS.1770 / EBU R128 — on the master bus post-limiter, playing the entire export region from start to finish, and reading the I value at the end of playback. This is distinct from relying on a real-time loudness display mid-session, which gives only a partial picture. Accurate integrated loudness requires a full-program pass; shortcutting to a 30-second sample introduces measurement error that accumulates when a track has a dynamic intro or a fading outro.
Diagram — Integrated Loudness: Diagram showing how momentary loudness blocks are gated and averaged to produce an integrated loudness reading, with platform targets marked.
Every integrated loudness — hardware or plugin — operates on the same core parameters. Know these and you can work with any implementation.
The primary delivery metric — a single LUFS value representing perceived loudness from first to last sample after dual gating. Streaming platforms compare this value against their normalization target (typically −14 to −16 LUFS) and apply gain accordingly. Mastered pop and electronic music typically lands between −8 and −14 LUFS; acoustic and classical content often sits between −16 and −23 LUFS.
Updates continuously during playback using a 3-second integration window, making it the most useful real-time display for monitoring vocal rides, mix section transitions, and bus level consistency. A well-balanced mix will show Short-Term values that hover within ±3 LU of the final Integrated reading for most of the program. Verses consistently 6+ LU below the chorus Short-Term value may indicate an energy gap that needs addressing in arrangement rather than mastering.
Computed over a 400 ms block and updated every 100 ms, Momentary loudness captures the loudest perceived instant in the program. It is useful for detecting aggressive transient spikes on kick drums, snare hits, and vocal consonants that may feel jarring even when the integrated reading is well-controlled. Most professional masters show Momentary peaks that sit 6–10 LU above the Integrated value; values more than 14 LU above Integrated suggest either highly dynamic content or a measurement artifact from a sparse opening section.
Defined in EBU Tech 3342, LRA measures the statistical spread between the 10th and 95th percentile of Short-Term loudness values after gating, expressed in LU. It is the most meaningful single-number descriptor of a track's dynamic character: heavily limited EDM and pop typically measures 3–7 LU, while cinematic scores and jazz recordings commonly show 12–20 LU. Mastering engineers use LRA alongside Integrated loudness to confirm that limiting has not destroyed musical dynamics — hitting −14 LUFS integrated with LRA above 8 LU is the hallmark of a dynamic, competitive master.
Measured by upsampling to at least 4× the program sample rate and detecting the highest reconstructed analog value, True Peak reveals clipping that standard sample-peak meters miss. The EBU R128 specification recommends a maximum True Peak of −1.0 dBTP for stereo delivery; Apple Music's Mastered for iTunes guidelines tighten this to −1.0 dBTP, and many engineers targeting lossy re-encoding use −1.5 dBTP for additional headroom. Exceeding 0 dBTP virtually guarantees audible distortion after AAC or MP3 encoding.
Set at −70 LUFS in the BS.1770 specification, the absolute gate prevents periods of silence, room tone, or audience noise from pulling the integrated reading down to unrealistically quiet values. For music production, this threshold is almost never reached during musically active sections, but it matters for tracks with long fades, spoken-word intros, or ambient field recordings where genuine silence is present. Engineers delivering podcast content or audiobooks to spoken-word platforms should verify that their content passes the absolute gate reliably.
Session-ready starting points. These values target Spotify, Apple Music, and YouTube normalization; always verify with a full-program loudness pass before delivery.
| Parameter | General | Drums | Vocals | Bass / Keys | Bus / Master |
|---|---|---|---|---|---|
| Target Integrated LUFS | −14 to −16 | −14 to −16 | −14 to −16 | −14 to −16 | −14 to −16 |
| True Peak Ceiling (dBTP) | −1.0 | −1.0 | −1.0 | −1.0 | −1.0 to −1.5 |
| Typical LRA (LU) | 5–10 | 4–7 | 6–10 | 5–9 | 4–12 |
| Headroom before limiter (dB) | 2–4 | 3–5 | 2–4 | 2–4 | 1–3 |
| Short-Term peak above I (LU) | 4–8 | 5–9 | 3–7 | 4–7 | 3–6 |
| Pre-master mix bus level (dBFS) | −6 to −3 | −8 to −5 | −8 to −6 | −7 to −5 | −3 to −1 |
| Classical / acoustic target (LUFS) | −23 | N/A | −20 to −18 | −20 to −18 | −20 to −23 |
These values target Spotify, Apple Music, and YouTube normalization; always verify with a full-program loudness pass before delivery.
The conceptual roots of integrated loudness measurement stretch back to the psychoacoustic work of Harvey Fletcher and Wilden Munson at Bell Labs, whose 1933 equal-loudness contours first mapped the frequency-dependent nature of human hearing sensitivity. For decades, broadcast and recording engineers relied on VU meters — standardized in 1939 by the combined effort of CBS, NBC, and Bell Labs — and later on PPM (Peak Programme Meters), standardized in IEC 60268-10 (1991). Both meter types measured signal amplitude rather than perceived loudness, leaving a persistent gap between what meters showed and what listeners heard. As digital audio production matured in the 1990s, the gap became a commercial liability: mastering engineers discovered that tracks perceived as louder tested better in focus groups, triggering an industry-wide race to maximize RMS levels by crushing dynamic range through hard limiting.
The formal scientific foundation for modern loudness measurement was laid by researchers at the Communications Research Centre Canada (CRC) and ITU Study Group 6. Researchers Glenn Soulodre and Michael Croghan published influential studies between 2000 and 2003 demonstrating that a specific combination of frequency weighting and gating could predict perceived program loudness with high accuracy across diverse content. Their work fed directly into the development of ITU-R BS.1770, first published in 2006, which defined the K-weighting filter, the mean square loudness computation, and the True Peak measurement methodology. The European Broadcasting Union adopted and extended this work in 2010 as EBU R128, adding the gating algorithm (contributed significantly by EBU researcher Florian Camerer and colleagues) and the Loudness Range metric, and mandating −23 LUFS integrated as the standard for European broadcast delivery.
The impact on broadcast was immediate and legally enforceable in several jurisdictions. In the United States, the Commercial Advertisement Loudness Mitigation (CALM) Act was signed into law in December 2010, directing the FCC to adopt rules — which it did in 2011 — requiring US television broadcasters to comply with ATSC A/85, an American loudness standard aligned with ITU-R BS.1770-3 and targeting −24 LUFS integrated for broadcast content. Viewers who had endured commercials dramatically louder than program content for decades finally got relief. For the music industry, the inflection point came when streaming platforms began implementing loudness normalization: Spotify introduced ReplayGain-style normalization around 2013 before transitioning to a BS.1770-based system targeting −14 LUFS; YouTube moved to −14 LUFS normalization for all content; Apple Music adopted −16 LUFS for its streaming normalization in alignment with its Mastered for iTunes (later Apple Digital Masters) program. By 2016, the loudness wars in commercial music were effectively over for normalization-aware engineers.
The evolution of the standard itself continued through successive revisions. ITU-R BS.1770-2 (2011) added multichannel dialogue-gating guidance; BS.1770-3 (2012) incorporated the gating algorithm from EBU R128 into the ITU specification; BS.1770-4 (2015) refined multichannel channel weighting. Today, virtually every professional loudness meter — hardware units from Dolby, hardware/software meters from Nugen Audio, Waves, TC Electronic, Orban, and the Beeper by Brecht De Man — implements BS.1770-4 in full. The LUFS unit itself, while mathematically identical to LKFS (Loudness, K-weighted, relative to Full Scale — the term used in ATSC A/85), has become the de facto standard label in music production contexts, a terminological preference that reflects the EBU influence over the broadcast one.
For mixing engineers, integrated loudness enters the workflow primarily as a sanity check before bouncing a mix for mastering. The goal at the mix stage is not to hit a streaming target — that is the mastering engineer's job — but to deliver a mix that has healthy dynamic range and sits at a level that gives the mastering chain sufficient headroom to work. A mix bus peak level of −6 to −3 dBFS with an integrated reading somewhere between −18 and −23 LUFS is a strong starting point. If the mix is already measuring −10 LUFS integrated before mastering, the mix bus limiter is doing too much work, and the mastering engineer will have little room to add punch, warmth, or air without pushing integrated loudness higher. Printing a mix loud is not the same as printing a mix competitive.
Mastering engineers use integrated loudness as the anchor point for the entire session. The workflow is systematic: import the mix, place a BS.1770-compliant meter post-chain, run the track from start to finish, note the integrated reading. Compare against the target for the primary platform (typically −14 LUFS for streaming-first releases). If the mix lands at −20 LUFS integrated, the mastering engineer knows they need to add roughly 6 dB of cumulative gain — through EQ, compression, parallel saturation, and a final true peak limiter — to reach the target. If it already reads −14 LUFS, the master either needs no additional limiting or the mix was already compromised by over-compression, which will reveal itself in a low LRA reading (below 5 LU for a genre that should have more dynamic movement). The integrated loudness number tells the mastering engineer where they are; LRA and true peak tell them how they got there and what the cost was.
Podcast producers and audio-for-video engineers use integrated loudness differently but with equal precision. The EBU R128 broadcast target of −23 LUFS is the standard for European television and a common reference for podcast delivery; the AES streaming media recommendation (AES TD1004) suggests −16 LUFS for podcast and streaming audio. Engineers working in these fields use loudness meters to normalize speech content — ensuring that episodes recorded in different environments, with different microphones and preamps, all arrive at the listener at a consistent perceived volume. This is particularly important in multi-host formats where one host's signal chain may be 4–6 dB louder than another's before normalization.
A more advanced application is using integrated loudness as a mix feedback tool during the creative phase. By regularly running a full-program loudness pass on a work-in-progress mix — not just glancing at a real-time meter — producers can track whether additions to the arrangement are lifting perceived loudness in a controlled way or whether the mix is drifting toward an over-compressed state. Plugins like iZotope Insight 2, Nugen VisLM, and the free DPAMETER by Dynameter display Integrated, Short-Term, Momentary, LRA, and True Peak simultaneously, giving producers a complete diagnostic picture. Making this a 10-minute checkpoint habit at the end of each session prevents the gradual loudness creep that causes mixes to sound exciting in the moment but exhausting on first play.
One email a week. The techniques behind the terms — curated by working producers, not algorithms.
Abstract knowledge becomes practical when you can hear it in music you know. These tracks demonstrate integrated loudness used intentionally, at specific moments, for specific purposes.
Streaming at approximately −14 LUFS integrated, 'bad guy' is a textbook example of how a minimalist arrangement can hit platform targets without aggressive limiting. The track's LRA measures around 7–8 LU — high for a mainstream pop single — because the dynamic contrast between the sparse verses and the punchier chorus sections is preserved rather than compressed out. FINNEAS's mix relies on surgical gain staging rather than a brick-wall limiter doing heavy lifting, meaning the kick and bass retain impact at normalized loudness. Listen on Spotify with normalization enabled alongside a heavily limited contemporary pop track: both arrive at the same perceived volume, but 'bad guy' sounds open and three-dimensional where the over-limited track sounds fatiguing.
HUMBLE. streams at approximately −9 to −10 LUFS integrated in its mastered form — significantly hotter than streaming normalization targets — meaning Spotify turns it down by roughly 4–5 dB on playback with normalization enabled. The production choice is deliberate: the mix was optimized for maximum punch on radio and unmanaged playback systems. At normalized loudness, it is competitive with more conservatively mastered tracks, but the original mastering decision captured a moment before platform normalization was universally enabled. Engineers studying this track should compare it in A/B against its normalized playback and observe what LRA compression at that level costs in terms of low-frequency punch on the 808s.
Mastered by Bob Ludwig, 'Get Lucky' sits around −12 LUFS integrated — a moderate level for a 2013 release that still required some gain reduction from streaming platforms. The track's LRA of approximately 9–10 LU is unusually high for a commercial electronic/disco record of its era, reflecting Ludwig's reputation for preserving dynamic character. The guitar transients from Nile Rodgers retain their pick attack through the master, and the dynamics between the instrumental sections and the full-band chorus are clearly audible. Use this track as an A/B reference for what a dynamically healthy commercial master sounds like after normalization.
From In Rainbows, mastered by Bob Ludwig, this track demonstrates deliberate dynamic architecture designed around the song's emotional arc. The sparse opening registers approximately −22 to −24 LUFS short-term; the full orchestral and electronic climax at around 3:00 hits approximately −8 to −9 LUFS short-term, creating a dynamic swing of 14+ LU across the program. The integrated reading lands around −15 LUFS, meaning streaming platforms require minimal gain adjustment. Studying the waveform envelope of this track alongside its loudness meter output reveals how the dual-gate algorithm handles the quiet sections — they are legitimately excluded from the integrated calculation, allowing the full dynamic of the arrangement to be preserved without misrepresenting the loudness of the active sections.
Targeting −23 LUFS integrated (EBU) or −24 LUFS (ATSC), broadcast loudness normalization is the most tightly specified implementation of BS.1770, with legal enforcement in many jurisdictions. Content delivered to broadcast chains must pass integrated loudness measurement, True Peak compliance (−1.0 dBTP), and often LRA constraints that limit dynamic range compression in favor of intelligibility. Hardware processors like the Jünger Audio Level Magic apply real-time loudness correction within these constraints for live broadcast chains where pre-mastering is not possible.
The dominant context for music producers today, streaming loudness normalization operates between −14 LUFS (Spotify, YouTube, Amazon Music HD) and −16 LUFS (Apple Music, Tidal), with platform-specific True Peak requirements. Unlike broadcast, streaming normalization is applied at the platform level rather than enforced at delivery — meaning over-loud masters are played back at reduced gain rather than rejected. The practical implication is that mastering to −14 LUFS with a healthy LRA and True Peak ceiling of −1.0 dBTP optimizes for the widest platform compatibility without sonic compromise.
Podcast loudness standards exist in a slightly less rigid space than broadcast. Spotify for Podcasters, Apple Podcasts, and most major podcast directories recommend between −16 and −19 LUFS integrated, with Spotify specifically recommending −14 LUFS and Apple recommending −16 LUFS. Spoken-word content benefits from dialogue normalization differently than music: the gating algorithm is especially important for removing silence between sentences that would otherwise deflate the integrated reading, and LRA values of 8–14 LU are common and desirable for natural-sounding speech dynamics.
Film and television post-production operates under SMPTE and Dolby standards targeting −27 LUFS (equivalent to the older Leq(m) reference widely used in cinema). This quiet reference level — 13 LU below the typical streaming music target — exists because film dialogue must remain intelligible over a wide dynamic range that includes loud action sequences and quiet dramatic moments. Composers delivering stems for film scoring workflows need to be aware that their music will be played back in this much quieter loudness context, meaning the dynamic architecture of a film score is designed around a very different loudness anchor than a commercial release.
DJ-optimized masters for club playback routinely sit between −6 and −9 LUFS integrated — far hotter than streaming targets — because DJ software normalization is typically defeated in professional club setups, and the DJ expects to have consistent gain structure across tracks in a set. Engineers mastering for vinyl cutting operate under different constraints entirely: cutting levels, frequency distribution, and phase coherence at low frequencies matter more than hitting a LUFS target. These contexts represent specialized cases where platform normalization is irrelevant, and mastering decisions are made for the specific playback environment rather than algorithmic normalization.
These MPW articles put integrated loudness into practice — specific techniques, real tools, and applied workflows.