CH-04 · Mastering
LUFS mastering standards
A finished master is judged twice: once by the ear and once by a meter. Loudness normalisation has made the second judgement decisive, because a delivery outlet will simply turn a loud master down and leave a quiet one alone. This article covers the units, the measurement, and the two targets that matter most in 2026.
Loudness units relative to full scale
Loudness units relative to full scale express perceived level, not peak level. A reading of -16 LUFS is quieter than -14 LUFS by two loudness units, whatever the peak values happen to be, and that relationship holds across piece of very different dynamic shape. The scale exists because a peak meter and the ear disagree: a heavily limited master can show a high peak while sounding no louder than a lightly handled one, and a sparse acoustic take can show a low peak while sounding quite forward.
The current measurement is defined jointly by EBU R128 and ITU-R BS.1770. Both describe a filter that shapes the signal roughly like the ear, a gating rule that ignores silence and very quiet passages, and an integration over the whole piece. The gating step is the part most often misunderstood. Without it, a long passage of room tone would drag the average down and the engineer would raise every remaining second to compensate. With it, the quiet sections are excluded and the reading describes the loud sections a listener actually remembers.
True peak
True peak measures the reconstructed analogue waveform rather than the stored samples. After conversion to a lossy delivery format, the waveform between samples can rise above the level suggested by the samples alone, so a file that looks safe on a sample meter can arrive clipped. The usual guard band is one decibel below full scale, written as -1 dBTP. That margin takes almost nothing in perceived loudness and prevents a whole class of delivery complaints.
Loudness and true peak must be read together. A master can sit exactly on its loudness target and still fail if its true peak sits above the ceiling; the fix is usually a small gain trim followed by fresh measurement, not more limiting.
Delivery targets in 2026
Two integrated targets dominate current practice. Spoken-word podcast delivery has settled at -16 LUFS integrated with a -1 dBTP ceiling, a pair of figures that leaves speech intelligible on small speakers without forcing the compressor to flatten every pause. Music delivery to streaming outlets has settled at -14 LUFS integrated, again with a -1 dBTP ceiling, because the outlet normalises every track to that point and returns anything louder at a reduced level.
| Deliverable | Integrated loudness | True peak ceiling | Dynamic aim |
|---|---|---|---|
| Spoken word podcast | -16 LUFS | -1 dBTP | 8 – 12 LU range |
| Music streaming | -14 LUFS | -1 dBTP | 6 – 10 LU range |
| Broadcast feature | -23 LUFS | -1 dBTP | 10 – 16 LU range |
| Archive access copy | -18 LUFS | -3 dBTP | 14 – 20 LU range |
The last row is the trap. A master at -9 LUFS is not louder to a listener once the outlet normalises it; it is only denser, with its peaks shaved and its transients softened, and on dynamic material it will lose more than it gains. Measuring before limiting is the discipline that keeps a delivery on target.
Measure integrated loudness over the whole piece, not over a section. A target hit in one chorus and missed in the next is not a finished master.