Someone put up this playlist on Youtube to test Youtube's Loudness Normalization settings.
Despite the wrong LUFS figures in the descriptions of the videos (they are ca. 1-2 LU/dB off) then it does proof that Youtube TURNS DOWN louder material to ca. -12.5 LUFS. Youtube does NOT turn up softer more dynamic material. I have a hunch that the uploader measured the whole song and put that number in the video description but then edited the audio to only 1 minute and therefore changing the integrated LUFS values for the file.
Here are my measurement figures for the playlist:
Video 1 (-24 LUFS): -25.2 LUFS
Video 2 (-17 LUFS): -18.9 LUFS
Video 3 (-11 LUFS): -12.9 LUFS
Video 4 (-6 LUFS): -12.4 LUFS
So it's only "Video 4 (-6 LUFS)" that has been loudness normalized to Youtube's target loudness. If video nr. 3 would have truly been -11 LUFS then it would have been turned down as well.
If softer more dynamic material would be turned up (without getting clipped) then it would have been possible to turn up videos nr. 1 and 2 (and possibly nr. 3 as well depending on the exact target /
algorithm)... but they are still untouched in level.
Example of a song that is -12.5 LUFS / PLR 11.5 ( peaking at -1 dBTP). Measured with Waveform Statistics in RX from iZotope and Dynameter from Meterplugs.
Russel Cottier measured the source files (downloaded from Youtube).
More on Youtube loudness over at Ian Shepherd's Production Advice page.
YouTube loudness normalisation – The Good, The Questions and The Problem
YouTube just put the final nail in the Loudness War’s coffin