I´ll chime in:
Before testing a lengthy DCP in a nearby cinema, I would like to check
if the audio levels are ok.
The feature is a 2hr documentary with 80% of dialogue (all Interviews), no narration.
I did only a very slight compression on the dialogue.
So it has dynamics.
I upmixed the sound to 5.1 from stereo.
But I did NOT use the original Stereo file but the stems
originally intended for Stereo. So the dialogue has its own stem/track.
I have no means to monitor in 5.1, so I still had to mix deaf.
The defunct Apple program Soundtrack allows to monitor in Stereo but deliver/ export5.1
Because the Dialogue it is not very clean (Room noise etc) ,
I decided to bleed 10% of the center it into the L and R channels.
This is why L/R have almost identical waveforms an follow the center about -20db.
(See first image) Still there is ambient (room) noise in L/R.
Ls/Rs are mixed down with a little reverb, only ambience and music.
No pan effects.
I made screenshots of:
A 5 Minute scene with dialogue only (again, 80% of the documentary are like this)
Just_Dialogue_5Min.jpg
A scene with about 25 seconds of dialogue followed by very loud music
that builds up (A chorale with orchestra from Bach)
Very_Loud_Music_Passage_2Min.jpg
A 5 minute scene with dialogue followed by loud music (not as loud as before) and dialogue again.
Dialogue_Loud_Music_Passage_5Min.jpg
But how do I read the data:
I.e. the integrated loudness in dialogue only:
Do I read the summation of
all channels?
So the center channel plus the 10% "bleed" in L/R
sums up to -35LUFS?
Using soft dialogue with loud music passages was my intention.
But I assume the dialogue is too soft, whereas the very loud music is about 3db too hot?
Best, Herbert