How to use it
- Choose an audio file. The measurement appears after decoding.
- Read the integrated loudness, the maxima, the loudness range and the peak.
- Pick a target. The gain and the resulting loudness and peak are shown.
- Keep the −1 dBFS peak ceiling ticked to avoid clipping, then save the normalized WAV.
How it works
Each channel is passed through the K-weighting filter of BS.1770 (a high shelf of about +4 dB above 1.7 kHz and a high-pass near 38 Hz) and squared. Integrated loudness averages 400 ms blocks that overlap by 75%: blocks below −70 LUFS are dropped, then blocks more than 10 LU below the average of the rest; the result is −0.691 + 10·log10 of the mean energy. Short-term uses 3-second windows and momentary 400 ms windows. Loudness range, after EBU Tech 3342, is the spread between the 10th and 95th percentiles of the short-term values, gated at −70 LUFS and 20 LU below their average.
Worked example
A full-scale 997 Hz sine in one channel measures −3.0 LUFS: its mean square is 0.5, the K-filter adds about +0.69 dB at that frequency, and −0.691 + 10·log10(0.5) + 0.69 ≈ −3.01. A voice recording at −20.0 LUFS with a peak of −3.0 dBFS, normalized to −14: the gain would be +6.0 dB, but the ceiling allows only +2.0 dB (peak −3 + 2 = −1), so the result is −18.0 LUFS and the tool says the ceiling limits the gain.
Tips
- Normalizing changes only the level, not the dynamics; reaching −14 with a quiet peak-heavy mix needs a limiter or compressor first.
- A large LRA (over about 15 LU) means very quiet and very loud parts; spoken word is usually much tighter.
- Pauses barely change the integrated value: blocks under −70 LUFS and far below the average are left out by the gates.