Selur's Little Message Board

Full Version: Method for Audio leveling or Normalize
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
First of all thanks for keeping updated this fantastic tool which i am starting to use to encode lossless Hi8 old tape.
However after encoding dozen of video with default settings i noticed some videos audio level are a bit all over which comes from their original recording levels back then.

I dig a bit in the Audio and Filtering options settings to find Gain control and normalize option however i don't understand how they work in my situation but find the help panel helpful.
- I did some encoding test on same file with/without auto gain option activated with default -n parameter
- I analyzed thoses videos in Audacity then Foobar2000 ReplayGain tool and some video have corrected audio level near 89db but some have same audio level before/after applying Auto gain
- Same for normalize some videos i see correction during Hybrid encodind some are not fixed.

Maybe someone could help me to figure which option i should set for my use case:
- I have 80+ captured lossless Lagarith files
- I encode them with x264 and QTGMC(Vapoursynth) with CRF18
- Audio AAC 192kbs

What audio filtering would you recommend me to activate? 
Already completed over 40 encoding without audio filtering, would i be able to normalize/fix audio afterward? i read that Regain can be amended via FFMPEG too, would it be the same as encoding them properly with gain control.
Disclaimer: I'm by no means an audiophile, nor do I have much experience with audio filtering. Audio filtering in the form it is there in Hybrid is because users asked about it. So I probably can't give great advice on it.
Personally I usually use 'Gain auto' + 'DRC Film Standard' for most stuff, but I suspect you want 'DRC None' and Normlize to a specific dB level,...

Hybrid uses sox for normalization&audio filtering.
'norm' and 'gain' are the option used for gain and normalization. (see: https://linux.die.net/man/1/sox)

Don't know why some files get normalized properly and others do not.

When doing audio normalization&co you might want to enable 'Conig->Input->Input Analysis->Check volumes', this will cause Hybrid to analysis the audio volumes with sox and show the results in the log, i.e.:
Code:
Detected audio volume values:
    Maximum amplitude: 0.119277
    Minimum amplitude: -0.157490
    Midline amplitude: -0.019107
    Mean norm: 0.003803
    Mean amplitude: 0.000001
    RMS amplitude: 0.007773
    Maximum delta: 0.097708
    Minimum delta: 0.000000
    Mean delta: 0.001382
    RMS delta: 0.002022
    Rough frequency: 1987
    Volume adjustment: 6.350
this might give some hints why some files get normalized as expected and others do not.

If audio gain&normalization usually requires a reencode of the audio, so if you want to adjust those with other tools, I would recommend not to highly re-compress the audio beforehand.

Cu Selur