How it works
Measuring loudness
Loudness follows ITU-R BS.1770-4, the method behind LUFS: each channel goes through a K-weighting filter that hears roughly as people do, the energy is taken over 400 ms blocks every 100 ms, blocks quieter than −70 LUFS are dropped, then blocks more than 10 LU under the average are dropped too, and what is left is averaged.
Peak is the largest sample in the file, in dBFS (0 is the loudest a file can store).
loudness = −0.691 + 10 × log10(mean energy of the gated blocks)
The gain
The gain is the target minus the measured loudness, applied to every sample. If that would push the peak past −1 dBFS, the gain is held back to keep 1 dB of headroom, and the tool says the result will be quieter than the target.
−14 LUFS is close to what the large music streaming services play at, −16 LUFS is a common podcast target, and −23 LUFS is the EBU R 128 broadcast level.
gain = min(target − loudness, −1 dBFS − peak)
The file you get
The output is 16-bit WAV at the source's sample rate and channel count, so nothing is lost to a second lossy encode. Convert it to MP3 or AAC afterwards if you need a smaller file.
