LUFS Loudness Reference

LUFS (Loudness Units Full Scale) and LKFS (Loudness, K-weighted, Full Scale) are identical measurements -- 1 LUFS equals exactly 1 LKFS. The difference is purely terminological: LKFS is the term used by the ITU-R BS.1770 standard, while LUFS is the term adopted by the EBU R 128 recommendation. Both apply K-weighting filters and gating algorithms to measure perceptual loudness relative to digital full scale. The industry uses both terms interchangeably.

Spotify normalizes to -14 LUFS by default (Normal setting). If your master is louder than -14 LUFS, Spotify applies gain reduction. If it is quieter, Spotify applies a limiter and gain increase. Spotify also offers Loud mode (-11 LUFS) and Quiet mode (-23 LUFS) for users. For best results, master to around -14 LUFS with a True Peak below -1.0 dBTP to avoid clipping artifacts after format conversion to OGG Vorbis or AAC.

True Peak (dBTP) measures the actual maximum waveform amplitude between digital samples using 4x oversampling interpolation, unlike standard digital peak meters that only check sample values. This matters because inter-sample peaks can exceed 0 dBFS during digital-to-analog conversion or lossy encoding (MP3, AAC, OGG), causing clipping distortion. Most standards require True Peak below -1.0 dBTP (EBU R 128, Apple, podcasts) or -2.0 dBTP (ATSC A/85, Netflix).

K-weighting is a two-stage frequency filter applied before loudness calculation. Stage 1 is a high-shelf pre-filter that boosts frequencies above 1681 Hz by approximately +4 dB, modeling the acoustic effect of the human head and ear shape on high-frequency perception. Stage 2 is an RLB (Revised Low-frequency B-curve) high-pass filter with -3 dB at 38 Hz, reducing low-frequency content that is less perceptually loud. This approach differs from A-weighting used in SPL measurements.

YouTube normalizes to -14 LUFS. If your content is louder than -14 LUFS, YouTube automatically applies gain reduction. However, since 2019, YouTube no longer applies gain increase to quieter content -- if your audio is below -14 LUFS, it stays at its original level. You can verify the applied adjustment through the "Stats for Nerds" overlay, where "content loudness" shows the gain change (e.g., -3.2 dB means 3.2 dB reduction from original). Recommended mastering range is -13 to -15 LUFS.

Netflix requires dialogue-anchored loudness at -27 LKFS with a tolerance of +/-2 LU, and True Peak below -2.0 dBTP. Amazon Prime targets -24 LKFS (+/-2 LU), Hulu targets -24 LKFS, and Apple TV+ targets -27 LKFS for Dolby content. All streaming video platforms use dialogue-anchored measurement rather than program-integrated loudness, because viewers adjust their volume based on dialogue intelligibility.

Use a two-pass approach with the loudnorm filter. Pass 1 (measurement): ffmpeg -i input.wav -af loudnorm=print_format=json -f null - which outputs measured_I, measured_TP, and measured_LRA values. Pass 2 (normalization): ffmpeg -i input.wav -af loudnorm=I=-14:TP=-1:LRA=11:measured_I=...:measured_TP=...:measured_LRA=... output.wav, inserting the measured values from Pass 1. This produces accurate loudness normalization to any target integrated loudness, True Peak, and LRA.

Loudness Range (LRA) is a statistical measure of dynamic variation in audio, calculated as the difference between the 10th and 95th percentile of short-term loudness values, expressed in LU (Loudness Units, where 1 LU = 1 dB). Low LRA (below 5 LU) indicates heavily compressed audio with little dynamic variation. High LRA (above 15 LU) indicates very dynamic content like classical music or film. Typical values are 5-10 LU for pop music, 8-15 LU for podcasts, and 15-25 LU for film/orchestral content.