/ˈlaʊdnəs/
Loudness is the perceived intensity of a sound, measured in LUFS (Loudness Units relative to Full Scale). It governs mastering targets, streaming normalization, and how energetic a mix feels to listeners.
Every producer has crushed a mix into a brick wall chasing loudness — and every one of them has watched it die on a streaming platform that quietly turned it back down. Understanding loudness means never making that trade again.
Loudness is the subjective perception of how intense a sound is, shaped by amplitude, frequency content, duration, and the nonlinear response of human hearing. It is distinct from level — a sine wave at −6 dBFS and a spectrally dense mix at −6 dBFS will not sound equally loud, because the ear integrates energy across time and weights different frequency bands unequally. This gap between digital level and perceived loudness is what drove the loudness war of the 1990s–2010s and what makes LUFS metering indispensable in contemporary production.
The modern standard unit for measuring loudness is LUFS — Loudness Units relative to Full Scale — defined by the ITU-R BS.1770 specification and its derivatives. An integrated LUFS measurement averages the loudness of an entire program, while short-term LUFS captures a three-second window and momentary LUFS reflects a 400-millisecond snapshot. These measurements model human auditory perception through a K-weighting filter: a gentle high-shelf boost above 2 kHz combined with a high-pass below 60 Hz, reflecting the ear's greater sensitivity to mid and high frequencies and its attenuation of deep low-end. True Peak measurement accompanies LUFS to catch inter-sample peaks that exceed 0 dBFS during D/A conversion, a common artifact of peak-normalized masters.
Streaming platforms have fundamentally redefined what loudness means in practice. Spotify normalizes playback to −14 LUFS integrated, Apple Music and Tidal to −16 LUFS, and YouTube to −14 LUFS. Masters louder than these targets are turned down in gain; masters quieter are left as-is or, on some platforms, turned up. This normalization architecture means that a master slammed to −7 LUFS integrated does not arrive at the listener louder than one at −14 LUFS — it arrives with roughly 7 dB of dynamic range destroyed and no competitive advantage. The incentive to over-limit has structurally collapsed. Producers and mastering engineers who understand this deliver masters with headroom, transient punch, and natural dynamic contrast that survive normalization and translate with authority.
Loudness interacts with every stage of the signal chain. At the tracking stage, consistent gain staging prevents accumulated distortion and gives processors the headroom they need to behave predictably. During mixing, loudness balance between elements — kick versus bass, vocal versus guitars — is the craft of level management and spectral carving. At mastering, integrated loudness is the final deliverable spec, alongside True Peak ceiling (typically −1 dBTP for streaming, −2 dBTP when encoding to lossy formats). A mastering engineer working to a brief of −14 LUFS integrated with a −1 dBTP ceiling has a concrete, measurable target, not an aesthetic judgment about whether something sounds loud enough.
Perceived loudness is also a compositional and arrangement tool. A track that sits at −16 LUFS with consistent energy and minimal dynamic variation may feel louder than a −14 LUFS track with large macro-dynamic swings between verse and chorus, because the brain adapts to the average intensity level. The relationship between statistical loudness (LUFS), peak level (dBFS), crest factor (the difference between the two), and perceived density is what separates engineers who understand loudness from those who simply chase numbers on a meter.
Loudness measurement under ITU-R BS.1770 begins with a K-weighting filter applied to each audio channel before energy summation. K-weighting consists of two stages: a high-shelf pre-filter with approximately +4 dB gain above 1.5 kHz (modeling the acoustic effect of the head on the eardrum) and a high-pass RLB filter rolling off below roughly 60 Hz. After K-weighting, the mean square of each channel is computed, channels are summed with a loudness coefficient (center and LFE channels receive different weights in multichannel formats), and the result is expressed on a logarithmic scale relative to full scale. Integrated LUFS applies a gating mechanism: blocks of audio more than 10 LU below the ungated average are excluded, preventing silent passages from dragging the measurement downward and giving a truer picture of the audible program material.
True Peak measurement addresses a specific failure mode of peak metering: inter-sample peaks. When a PCM signal is converted back to analog, the reconstruction filter interpolates between sample values. It is entirely possible for the interpolated waveform between two samples to exceed 0 dBFS even when no individual sample does — an inter-sample peak. This is particularly common in content with significant high-frequency energy and after lossy encoding (AAC, MP3), where codec artifacts can add 1–3 dB of transient energy. True Peak meters oversample the signal (typically 4× to 8×) to detect and display these intersample excursions. A True Peak ceiling of −1 dBTP gives 1 dB of margin for codec-induced overshoot; −2 dBTP is conservative and appropriate for maximum streaming platform safety.
The relationship between LUFS and perceived loudness involves several psychoacoustic effects beyond pure amplitude. The Fletcher-Munson equal-loudness contours (updated as ISO 226:2003) show that human hearing is most sensitive between approximately 2 kHz and 5 kHz, and significantly less sensitive at low frequencies at moderate listening levels. A mix with substantial energy in the 2–5 kHz presence region will measure louder perceptually than its LUFS reading might suggest relative to a bass-heavy mix at the same integrated measurement. Spectral loudness balance — the distribution of energy across the frequency spectrum — is therefore as important as the aggregate loudness number. Producers mastering for electronic music, where sub-bass content is heavy, may find that hitting −14 LUFS integrated while maintaining low-end power requires more aggressive high-frequency limiting control or multiband management than a guitar-forward mix at the same target.
Dynamic range and loudness are inversely related by the physics of headroom. A signal with a crest factor of 14 dB (the difference between its True Peak and its integrated loudness) has significant transient headroom and will sound punchy and three-dimensional. Reducing crest factor to 6 dB through hard limiting raises integrated loudness while compressing micro-dynamic detail. Modern loudness-normalized streaming environments reward higher crest factors: a master at −14 LUFS with a 12 dB crest factor and a master at −7 LUFS with a 5 dB crest factor will be played back at nearly identical perceived volumes, but the former preserves transient character that translates to listener engagement, emotional punch, and, critically, better performance in A/B comparisons on high-resolution playback systems.
Short-term and momentary LUFS readings are the working tools of a mix session, not just mastering. Momentary LUFS on individual buses — kick, bass, vocals — shows relative contribution to the mix's overall loudness in real time. Engineers targeting a final integrated master of −14 LUFS typically work with a mix that peaks short-term between −12 and −10 LUFS at its loudest passages, leaving 2–4 LU of mastering headroom for limiting and saturation without over-compression. This working headroom is the practical link between mix loudness management and the final deliverable specification.
Diagram — Loudness: Diagram comparing an over-limited brick-wall master versus a dynamic master at the same streaming playback level, showing crest factor, LUFS targets, and True Peak ceiling.
Every loudness — hardware or plugin — operates on the same core parameters. Know these and you can work with any implementation.
Integrated LUFS is the primary deliverable specification for mastering to streaming platforms. It time-averages K-weighted loudness across the entire track, excluding gated (near-silent) passages. Target −14 LUFS for Spotify and YouTube, −16 LUFS for Apple Music and Tidal, −23 LUFS for broadcast (EBU R128). Submitting a master at −7 LUFS integrated results in roughly 7 dB of gain reduction on playback — identical loudness, destroyed dynamics.
True Peak detects amplitude excursions between PCM samples by oversampling (typically 4×–8×) the signal and examining the reconstructed waveform. The standard ceiling for streaming is −1 dBTP; for lossy codec delivery (MP3, AAC) use −2 dBTP to absorb encoder-induced overshoot of 0.5–1.5 dB. Exceeding 0 dBTP causes clipping in the D/A converter and audible distortion on consumer playback, even when no individual PCM sample clips.
Short-term LUFS measures the K-weighted loudness of the most recent 3 seconds, updating every 100 ms. It reveals loudness variation between sections — verse, chorus, breakdown — and is the most useful meter for real-time mix work. A well-structured mix targeting −14 LUFS integrated typically shows short-term values of −10 to −12 LUFS at the loudest chorus and −18 to −20 LUFS in quiet intros, providing macro-dynamic contrast that survives normalization.
Momentary LUFS uses a 400 ms integration window, fast enough to track individual elements like kick drum hits and vocal phrases. It is not gated. Use momentary readings on individual channel buses to understand each element's contribution to mix loudness in perceptual terms. A kick drum with momentary peaks of −10 LUFS while the mix sits at −18 LUFS short-term will dominate perceived loudness; identifying this allows surgical gain or dynamic correction rather than blanket level adjustment.
Crest factor is calculated as True Peak (dBTP) minus integrated loudness (LUFS), expressed in LU or dB. A crest factor of 12–14 LU indicates a dynamic, punchy master. Values below 6 LU characterize heavily limited, brick-wall masters typical of the loudness war era. In a streaming environment where loudness is normalized, maximizing crest factor within your LUFS target is the single most impactful technique for preserving transient energy, depth, and emotional dynamics.
Loudness Range (EBU Tech 3342) describes the statistical spread of short-term loudness values over a program, expressed in LU. Commercial music typically measures LRA 4–9 LU; classical recordings may reach 20+ LU. An LRA below 4 LU signals a flat, over-compressed master regardless of its LUFS reading. Broadcast specifications (EBU R128) often specify maximum LRA alongside target integrated loudness to prevent both over-compression and excessive dynamics.
Session-ready starting points. Values assume streaming delivery (Spotify, YouTube, Apple Music); broadcast targets differ significantly — use EBU R128 (−23 LUFS) for television and podcast distribution.
| Parameter | General | Drums | Vocals | Bass / Keys | Bus / Master |
|---|---|---|---|---|---|
| Target Integrated LUFS | −14 LUFS | −14 LUFS | −16 LUFS | −14 LUFS | −14 to −16 LUFS |
| True Peak Ceiling | −1 dBTP | −1 dBTP | −2 dBTP | −1 dBTP | −1 dBTP |
| Typical Short-Term LUFS (loud section) | −10 to −12 LUFS | −9 to −11 LUFS | −12 to −14 LUFS | −11 to −13 LUFS | −10 to −12 LUFS |
| Recommended Crest Factor | 8–14 LU | 10–16 LU | 10–18 LU | 8–14 LU | 8–12 LU |
| Loudness Range (LRA) target | 5–9 LU | 4–8 LU | 6–10 LU | 4–7 LU | 5–9 LU |
| Mix headroom before mastering | −6 to −3 dBFS peak | −6 to −4 dBFS peak | −6 to −3 dBFS peak | −6 to −4 dBFS peak | −6 to −3 dBFS peak |
| Gain reduction on limiter at master | 2–4 dB GR | 1–3 dB GR | 2–5 dB GR | 2–4 dB GR | 2–5 dB GR |
Values assume streaming delivery (Spotify, YouTube, Apple Music); broadcast targets differ significantly — use EBU R128 (−23 LUFS) for television and podcast distribution.
The story of loudness in recorded music is inseparable from the history of vinyl mastering. From the late 1940s through the 1970s, loudness was a physical constraint: the wider and deeper a groove was cut, the louder the pressing, but wider grooves took up more space on the disc and limited playing time. Engineers like Bob Clearmountain and Bob Ludwig worked within strict mechanical limits. The RIAA equalization curve, standardized in 1954, defined the playback characteristic and required complementary pre-emphasis at cutting — another factor shaping what loud could mean on wax. These constraints naturally preserved dynamic range, because the medium demanded it.
The arrival of the Compact Disc in 1982 removed the physical ceiling. With 16-bit PCM offering a theoretical 96 dB of dynamic range and a fixed digital ceiling at 0 dBFS, there was no groove depth or tape saturation to limit ambition. Initially, engineers were conservative — early CD masters often sounded quieter than vinyl equivalents. But by the late 1980s, mastering engineers including Bob Ludwig, Ted Jensen at Sterling Sound, and Greg Calbi began discovering that limiters and multiband compressors could raise the average level of a CD master without exceeding 0 dBFS, making it sound louder than competing releases on first impression. Radio programmers and A&R executives interpreted this perceived loudness as energy and commercial viability, and the arms race began.
The loudness war accelerated dramatically through the 1990s and 2000s. Oasis's 1995 album (What's the Story) Morning Glory? was mastered at levels that drew comment from engineers at the time. By 2008, Metallica's Death Magnetic — mastered by Ted Jensen under label pressure — became a cultural flashpoint: the Guitar Hero rip of the same songs, which had not been hard-limited, sounded dramatically better, and an online petition gathered tens of thousands of signatures demanding a remaster. Researcher and audio engineer Bob Katz had been documenting the trend since the late 1990s, publishing loudness data showing that average integrated loudness of pop masters had climbed from roughly −20 LUFS in the early CD era to −8 LUFS by the late 2000s — a 12 LU increase in perceived intensity bought entirely through dynamic range destruction.
The regulatory and technical response came through broadcast first. The EBU (European Broadcasting Union) published EBU R128 in 2010, standardizing integrated loudness for broadcast at −23 LUFS with a True Peak ceiling of −1 dBTP and a maximum LRA. The ITU-R BS.1770 measurement standard, revised through versions BS.1770-3 and BS.1770-4, formalized the K-weighting algorithm and gating methodology that underpins all modern loudness measurement. The United States followed with the CALM Act (Commercial Advertisement Loudness Mitigation Act), signed into law in 2010 and enforced from 2012, requiring television broadcasters to match advertisement loudness to program loudness. These regulatory actions established the measurement infrastructure that streaming platforms would later adopt.
Streaming normalization changed the commercial incentive structure permanently. Apple's SoundCheck feature, present in iTunes since 2001, was an early implementation of playback-level normalization but had limited adoption. Spotify introduced loudness normalization in 2013, initially at −11 LUFS, and revised its target to −14 LUFS in 2017 following lobbying from mastering engineers including Ian Shepherd and Thomas Lund, who had been publicly advocating for dynamic-friendly targets. Apple Music, YouTube, TIDAL, and Amazon Music all followed with their own normalization implementations, all clustering between −14 and −16 LUFS integrated. By the early 2020s, the loudness war was effectively over for streaming delivery. The remaining battleground is sync licensing and physical media, where some clients still request louder masters — though even here, attitudes have shifted markedly among engineers who understand the psychoacoustics.
In the mix session, loudness management begins at gain staging. The goal is to keep individual channels, buses, and the stereo sum at consistent operating levels — typically with the stereo mix bus peaking between −6 and −3 dBFS — so that the mix's perceived loudness is controlled by deliberate level and dynamic decisions, not by accumulated gain from poorly trimmed channels. Checking the mix's short-term LUFS reading at the loudest section gives an immediate sense of how much headroom remains for mastering. If the loudest chorus already reads −9 LUFS short-term with peaks at −1 dBFS, the master is going to require significant limiting to reach −14 LUFS integrated, and that limiting will cost dynamic range. Building the mix with 3–5 dB more headroom — peaks at −4 to −6 dBFS, loudest section short-term around −12 LUFS — gives the mastering stage room to breathe.
Drum and percussion production has a direct relationship with loudness management. Kick and snare transients are the primary source of a mix's crest factor: a tight, punchy kick that hits −6 dBFS peak while the mix idles around −18 LUFS short-term gives a crest factor close to 12 LU. Transient shapers, parallel compression, and attack/release tuning on bus compression all interact with how much of this crest factor survives into the master. A drum bus running −3 to −4 dB of gain reduction on a Neve-style compressor (API 2500, SSL G-Comp, or plugin equivalents) is shaping the loudness profile of the entire mix; the kick-to-mix relationship is a loudness relationship as much as a level one.
Vocal loudness is perceived disproportionately because the ear is most sensitive in the 2–5 kHz presence range where vocals dominate. A vocal that measures the same LUFS as the drums will be perceived as louder because of its spectral placement. Engineers working in pop and R&B typically ride vocals using automation to keep short-term loudness consistent within ±2 LU through the song, then use a combination of optical and VCA compression (UA LA-2A, SSL channel compressor, or their plugin equivalents) to catch dynamic variation that automation doesn't address. The goal is a vocal that sits consistently in the loudness hierarchy of the mix without requiring constant level automation.
Bass and low-end management affects both the measured and perceived loudness of a master in complex ways. Sub-bass below 60 Hz is largely attenuated by K-weighting, meaning heavy sub-bass energy contributes relatively little to LUFS readings. This creates a trap: a mix with massive sub content may measure at −14 LUFS integrated but sound dramatically louder on systems with good bass extension, and significantly quieter on small speakers. High-passing the sub-bass at 30–40 Hz and using multiband limiting to control the 60–120 Hz region allows the low end to hit hard on capable systems while keeping the loudness measurement honest and predictable. On the master bus, limiting the sub-bass separately from the mid-range prevents low-end transients (kick, bass) from triggering gain reduction that compresses the entire mix.
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 loudness used intentionally, at specific moments, for specific purposes.
Get Lucky measures approximately −11 LUFS integrated — moderately loud by 2013 standards but with a noticeably higher crest factor than its contemporaries, around 10–11 LU. The interplay of Nile Rodgers's Stratocaster and the tightly compressed live drums creates a mix whose perceived density is achieved through arrangement and spectral balance rather than peak limiting. Listen at the chorus entry (2:20): despite substantial energy, individual transients — the snare, the guitar stabs — remain distinct and punchy. This was the mastering aesthetic of Chilly Gonzales and the Columbia/Sony mastering team honoring the dynamic range of the Wendy Carlos/Quincy Jones-influenced production rather than flattening it for radio competition.
Bad guy was widely noted at release for its unusually restrained loudness — approximately −14 LUFS integrated — at a time when most pop masters were pushing −7 to −9 LUFS. Finneas mixed and mastered primarily on Yamaha HS5 nearfields and Apple AirPods, prioritizing translation across small speakers. The production's psychoacoustic loudness comes from extreme spectral contrast: sub-bass energy below 80 Hz, near-silence in the upper-bass and low-mids, and sharp presence in the vocal 3–5 kHz range. On streaming platforms, where normalization plays it at its native level, the track's perception of intimacy and menacing density demonstrates that measured loudness and perceived impact are separable creative dimensions.
HUMBLE. opens with a brief classical piano passage before the beat drops — a deliberate macro-dynamic contrast that demonstrates loudness as an arrangement tool. The track's integrated LUFS is approximately −8 to −9 LUFS, reflecting the aggressive limiting typical of hip-hop masters of the era, but the opening 17 seconds creates a perceived loudness differential of 12+ LU that makes the drop physically impactful beyond what the limiting alone achieves. This technique — using brief quiet sections to reset the listener's loudness adaptation before a dense drop — is a compositional loudness strategy that works independently of metered values. At the 0:18 drop, monitor the sub-bass bloom and the 808 transient against the snare for the crest factor relationship that defines the track's punch.
Blinding Lights measures approximately −7 to −8 LUFS integrated — a hard-limited master that sacrifices approximately 6–7 LU of crest factor for competitive loudness in the CD and terrestrial radio context. On streaming platforms that normalize to −14 LUFS, it is turned down 6–7 dB, arriving at the listener with compressed transients but a densely saturated spectral texture. The mastering, handled at Metropolis Studios London, uses the extreme low-end tightness and midrange saturation of the synthwave influence to maintain perceived energy even after normalization gain reduction. It is instructive to A/B this track against Billie Eilish's output — both normalize to the same playback level, but the Weeknd's crest factor compression is audible in transient smear on the snare and the flattened depth of the synthesizer pad reverb tails.
Integrated loudness is the time-averaged, K-weighted loudness measurement across the full duration of a program with gating applied. It is the primary deliverable metric for streaming and broadcast mastering. Hardware meters like the TC Electronic LM2n display integrated LUFS in real time and are standard in broadcast mastering facilities; the Dorrough 280-A was widely used in broadcast production suites during the transition from peak to loudness-based standards. Integrated loudness answers the question: how loud is this program overall?
Short-term loudness uses a 3-second integration window to show loudness variation across sections of a program, updated every 100 ms. It is the most practical meter for mix work, revealing verse-to-chorus dynamics, the loudness contribution of individual sections, and the headroom available for mastering. The Nugen Audio VisLM has been a standard in post-production and mastering for its graphical history display of short-term, momentary, and integrated values simultaneously.
Momentary loudness integrates over 400 ms without gating, capturing the loudness of individual musical events — a kick drum, a vocal phrase, a snare hit. It is fast enough to be useful on individual channel inserts for balancing elements by perceived loudness rather than peak level. iZotope Insight 2 displays momentary LUFS alongside a loudness histogram and spectrogram, making it a comprehensive monitoring solution for both mix and mastering engineers who need to correlate spectral content with loudness readings.
Perceived loudness is the subjective experience of intensity shaped by spectral content, temporal integration, and auditory masking — it cannot be captured by any single meter value. Equal-loudness contours, Fletcher-Munson curves, and psychoacoustic models inform how engineers make mixes feel loud by boosting presence frequencies (2–5 kHz), controlling dynamic variation, and using saturation to increase spectral density without raising peak levels. The Orban loudness processors used in broadcast since the 1980s model aspects of perceived loudness through multiband processing that optimizes subjective impact within technical limits.
True Peak is not loudness in the psychoacoustic sense but the essential technical companion to LUFS measurement — the ceiling that prevents clipping during D/A conversion and codec processing. True Peak meters oversample the digital signal at 4× to 8× and detect inter-sample excursions. The TC Electronic Clarity M hardware meter is used in broadcast and streaming QC workflows for its simultaneous display of integrated LUFS and True Peak with delivery-spec compliance indicators.
These MPW articles put loudness into practice — specific techniques, real tools, and applied workflows.