el 2012-05-20 a las 21:48 Nikos Chantziaras escribió: > On 20/05/12 12:41, Jesús J. Guerrero Botella wrote: > > Just for sake of correctness, what the op wants is called > > normalization, in the world of sound edition. > > Actually, no. That's not what he wants. Normalization simply adjusts > to 0db. How loud something sounds however is not a simple matter of > what the maximum peak of a waveform is. ReplayGain actually analyzes > the music to tell how loud it *sounds*, not how loud it actually is.
[...] > Normalization makes audio equally loud for hardware. ReplayGain makes > audio equally loud for humans. :-) actually, that isn't quite correct either... that's not the difference between normalization and replaygain, you are mixing different things. - normalization is process that modifies all the data in a file to adjust it to a reference level. as such, it only works in uncompressed audio; - replaygain is an algorithm that tries to estimate the perceived loudness of a sound file, and calculates the gain level needed during playback (hence the name) to adjust it to a reference loudness level. this gain level is written in the metadata of the file (not all file formats support it), and has to be understood by the playback device. it does not modify the actual audio data (that is, it does not normalize). now, normalization does not "simply adjust to 0dB", you can of course normalize to whatever level you want (usually, *not* 0dB). moreover, normalization doesn't necessarily mean peak level normalization, there's also loudness normalization. RMS normalization its most basic form of loudness normalization, but there are more complex algorithms that take into account the response curve of the human ear (like the replaygain algorithm).