什麼是 LUFS？面向音乐製作人的響度詳解

LUFS 實際測量什麼

LUFS（Loudness Units relative to Full Scale）是衡量感知響度的國際標準單位。與簡單的峰值測量不同，LUFS 考慮了人耳對不同頻率的敏感度，提供了更準確的響度感知測量。

LUFS 是所有主要流媒體平臺（Spotify、Apple Music、YouTube、Tidal）用來標準化播放的標準。理解 LUFS 對於確保你的音樂在所有平臺上表現一致至關重要。

原因 LUFS Replaced Peak Metering

Before LUFS, engineers mixed to peak dBFS — the instantaneous maximum sample value. The problem: two tracks can share an identical peak of −0.1 dBFS yet sound wildly different in loudness. A heavily compressed pop master with average energy near the ceiling sounds ear-splitting; a lightly compressed orchestral recording with occasional peaks at the same ceiling sounds quiet in between. Listeners constantly reached for the volume knob, and the loudness war escalated as mastering engineers drove integrated levels higher to sound louder on radio.

LUFS solved this by measuring average perceived energy across time, gated to exclude silence. Streaming platforms adopted it to enforce a level playing field: every track gets played back at the same perceived loudness regardless of how hard it was mastered.

LUFS 的型別

A loudness meter displays three distinct LUFS values simultaneously. Knowing which one to watch — and when — is the difference between a useful reading and a meaningless one.

• Integrated LUFS (I) The average loudness from the start of the measurement to the end, gated per EBU R 128^[2] to ignore passages more than 10 LU below the running average. This is the number streaming platforms compare against their reference level. It is the most important figure for mastering decisions.
• Short-Term LUFS (S) A 3-second sliding window average, updated every 100 ms. Useful for monitoring the loudness of specific sections — verses, choruses, drops — in real time during playback. It shows you where the loud and quiet moments are within a mix without waiting for the integrated average to settle.
• Momentary LUFS (M) A 400 ms window, updated every 100 ms. Shows transient peaks in loudness — a snare hit, a sudden vocal peak, a kick drum. Mostly useful for broadcast compliance and dialogue loudness, though producers use it to gauge how punchy individual hits read relative to the rest of the material.
• Loudness Range (LRA) Not a LUFS figure itself, but displayed alongside it on most meters. LRA is the difference in LU (loudness units) between the quieter and louder parts of the programme, excluding outlier extremes. A wide LRA means high dynamic range; a narrow LRA means heavy compression. A typical pop master might have an LRA of 4–8 LU; a classical recording can exceed 20 LU.

流媒體平臺目標

Sample-peak meters read the highest sample value in the digital file. But when an audio file is decoded and played back — especially after lossy encoding to AAC or MP3 — the waveform is reconstructed, and the actual analog-domain peak can exceed the highest sample value. These are called inter-sample peaks.

True peak measurement, specified in ITU-R BS.1770^[1], oversamples the signal (typically 4× or higher) to detect these in-between peaks before encoding. The result is reported as dBTP (decibels, True Peak).

Most distribution guidance recommends a maximum true peak of −1 dBTP. This 1 dB of headroom prevents inter-sample peaks from causing audible distortion after the platform's encoder compresses the file. If you are delivering high-resolution lossless files (for example, to services that offer HD tiers), some engineers target −2 dBTP for additional codec headroom.

如何測量 LUFS

Three measurement systems appear on different meters in a DAW. They serve different purposes and should not be confused.

How Loudness Normalization Works on Streaming Platforms

When you upload a track to a streaming service, the platform's encoder measures the integrated LUFS of your file. At playback, it applies a gain offset to bring that track to the platform's reference level. Your audio file is not re-encoded; the volume adjustment happens transparently in the playback stream.

The practical consequence: mastering louder than the platform target delivers zero advantage. If Spotify normalizes to −14 LUFS^[3] and you deliver a master at −7 LUFS, Spotify turns it down 7 dB. Your listener hears it at the same loudness as everything else — but now your transients are crushed and your mix has lost all the dynamic breathing room you worked to achieve.

Conversely, a track mastered at −14 LUFS or quieter plays at its native level (or close to it). The dynamics are intact. The mix sounds the way it was intended.

Streaming Platform Loudness Reference Levels

Each platform publishes (or has documented through its own tools) an integrated LUFS reference level. The table below covers only platforms with a confirmed primary source at time of writing.

Note on other platforms: Apple Music, YouTube, Tidal, and Amazon Music are widely reported by mastering engineers and industry publications to normalize in the −14 to −16 LUFS range, but as of this writing, none of those platforms publish a dedicated primary-source help article stating their exact integrated LUFS reference level. The information circulating in production blogs and forums is consistent but derived from engineer testing, not official documentation. If you distribute across multiple platforms, −14 LUFS integrated with a −1 dBTP true peak ceiling is the practical industry default that will clear every major service without loudness penalty.

What LUFS Should You Master To? (And 原因 −14 Isn't Always Right)

The short answer repeated everywhere is “master to −14 LUFS for streaming.” The nuance is more useful than the slogan.

1. Understand what −14 LUFS actually means for your genre
   A heavily side-chained techno track with near-constant energy will land near −14 LUFS integrated with modest compression. An acoustic folk track with quiet verses and loud choruses might already average −16 to −18 LUFS when properly mixed — and that is fine. Spotify will not turn it up above −14 LUFS[3], so a quiet track stays quiet — but if the mix sounds balanced and natural, that is not a problem worth fixing.
2. Do not limit your master hard just to hit −14 LUFS
   Over-limiting to reach a loudness target kills transients and introduces distortion. A master at −11 LUFS that required harsh limiting sounds worse than a −15 LUFS master with clean dynamics — and the louder one will just be turned down anyway. Master until it sounds right, then check the integrated reading.
3. Use −14 LUFS as a ceiling, not a floor
   Think of −14 LUFS as the point beyond which streaming normalization starts working against you. Anything louder gets attenuated. At −14 LUFS or below, your track plays as delivered. Shoot for a range of −14 to −16 LUFS for most genres; −16 to −18 LUFS for dynamic or acoustic music.
4. Set a hard true peak limiter at −1 dBTP
   This is non-negotiable regardless of your integrated LUFS target. True peaks above −1 dBTP cause clipping artifacts after the platform's AAC or Ogg Vorbis encoder runs on your file. Place a true peak limiter as the last processor in your mastering chain.
5. Check the integrated reading on the final rendered file, not during playback
   In DAWs like Ableton and FL Studio, the integrated reading continues to climb during silence at the start or end of a session. Render the final stereo file and measure it in a standalone loudness meter to get an accurate integrated LUFS reading.

如何製作 Measure LUFS: Free Meters

Every major DAW now ships with some form of loudness metering, but standalone meters give you the cleanest workflow for final checks.