MeGUI - Audio too loud

[wuemura]

When I was using analog audio on my computer I've nerver mind the audio volume of the videos encoded with MeGUI, but after switch to digital connection with my HT I've noticed that the audio is too loud and with some saturation.

Since I use different sources to encode is hard to guess how to propper normalise the audio, my videos are encoded with x264 DXVA-HD-HQ + Nero AAC LC 64Kbps.

Normally what I do now is to use "wavegain" with the follow command:

wavi.exe job1.avs AUDIO.wav
Found PCM audio: 2 channels, 48000 Hz, 16 bits, 346.680000 seconds.
Writing WAV file "AUDIO.wav"...
WAV file written successfully.

wavegain -l -f audio.log -y AUDIO.wav
 Analyzing...

    Gain   |  Peak  | Scale | New Peak |Left DC|Right DC| Track
           |        |       |          |Offset | Offset |
 --------------------------------------------------------------
  +4.39 dB |  12211 |  1.66 |    20241 |    0  |     0  | AUDIO.wav

 Applying Gain of +4.39 dB to file: AUDIO.wav

 WaveGain Processing completed normally

neroAacEnc.exe -cbr 64000 -lc -if AUDIO.wav -of "AUDIO.m4a"

del AUDIO.wav

If MeGUI could use this tool to make it "automatic" gain users could decide to use it or not.

Thanks.

[tebasuna51]

WaveGain don't support multichannel audio.
You have the 'Normalize Peaks to ...' to do the same job

[wuemura]

For multichannel materials you still can use 'peaks'.
Normalizing audio by 'peaks' you can get errors with some audio materials that need reduction.
The idea is to add an option for automatic normalization with stereo audio.

[tebasuna51]

Originally Posted by wuemura

...
Normalizing audio by 'peaks' you can get errors with some audio materials that need reduction
....

The method is scan all the audio to reach the max audio value and, in a second pass, amplify until the value selected.
Only if you select values over 100% you can get errors.

[wuemura]

Originally Posted by tebasuna51

The method is scan all the audio to reach the max audio value and, in a second pass, amplify until the value selected.
Only if you select values over 100% you can get errors.

But how can you, be sure, if '100%' is always what you need?
Boosting levels to 0dB might not make the audio clip, but don't forget that it will if it get processed or filtered by EQ for example, that is exactly what is happening with the defaul value of 100%.

The problem is that the tool is not smart enought to decide the propper gain value, instead of a fixed 100% peak.

You might read something called intersample peaks:
http://en.wikipedia.org/wiki/DBFS

[tebasuna51]

Originally Posted by wuemura

But how can you, be sure, if '100%' is always what you need?

I'm sure, but if you aren't you can put the desired value.

The problem is that the tool is not smart enought to decide the propper gain value, instead of a fixed 100% peak.

I think is enough smart and introduce more sophisticated tools is not justified.

You might read something called intersample peaks:
http://en.wikipedia.org/wiki/DBFS

The improbable intersample peaks can occur at very high frequencies and volume. And I think you never can ear any clip in your encoded AAC LC 64Kbps, these high frequencies are cut off.

[wuemura]

Originally Posted by tebasuna51

I'm sure, but if you aren't you can put the desired value.

What is best?
What I desire or what should be done properly?

Originally Posted by tebasuna51

I think is enough smart and introduce more sophisticated tools is not justified.

Is justifiable when the previous return erros, anyone can test and see for it self.

Originally Posted by tebasuna51

The improbable intersample peaks can occur at very high frequencies and volume. And I think you never can ear any clip in your encoded AAC LC 64Kbps, these high frequencies are cut off.

If you use a $10 speaker at your desktop you will never hear it, in that point you are right. But things go bad when you feed an audio 'peaked' by MeGUI in to an EQ to hear how saturated and clipped it is.

And in cases like this a more sofisticated software as mentioned before, it is, justifiable.

[tebasuna51]

Originally Posted by wuemura

... But things go bad when you feed an audio 'peaked' by MeGUI in to an EQ to hear how saturated and clipped it is.

Please put a sample (source and 'peaked').

[wuemura]

Originally Posted by tebasuna51

Please put a sample (source and 'peaked').

Here it is:
http://www.4shared.com/file/14747352...les-MeGUI.html

Original sample.avs script:

DirectShowSource("E:\video.avi")
Trim (191,1090).fadeout(5)

Extracted with command:

wavi.exe sample.avs AUDIO - Original.wav
Found PCM audio: 2 channels, 48000 Hz, 16 bits, 30.063375 seconds.
Writing WAV file "AUDIO - Original.wav"...
WAV file written successfully.

AUDIO - Original.wav is the source.
AUDIO - MeGUI.mp4 is the source encoded with MeGUI with 'peak' at 100% with profile "Nero AAC LC 64Kbps".
AUDIO - wavegain.mp4 is the source normalized with wavegain and encoded with neroAacEnc.

This are the command used to make the last sample:

wavegain

wavegain -l -f audio.log -y AUDIO - wavegain.wav
 Analyzing...

    Gain   |  Peak  | Scale | New Peak |Left DC|Right DC| Track
           |        |       |          |Offset | Offset |
 --------------------------------------------------------------
  +2.30 dB |  12089 |  1.30 |    15754 |    0  |     0  | AUDIO.wav

 Applying Gain of +2.30 dB to file: AUDIO - wavegain.wav

 WaveGain Processing completed normally

neroAacEnc

neroAacEnc.exe -cbr 64000 -lc -if "AUDIO - wavegain.wav" -of "AUDIO - wavegain.mp4"
*************************************************************
*                                                           *
*  Nero AAC Encoder                                         *
*  Copyright 2008 Nero AG                                   *
*  All Rights Reserved Worldwide                            *
*                                                           *
*  Package build date: Sep 17 2008                          *
*  Package version:    1.3.3.0                              *
*                                                           *
*  See -help for a complete list of available parameters.   *
*                                                           *
*************************************************************

Processed 29 seconds...

Pay close attention to actors voice and near the end of the file processed by MeGUI, now play the file on your HT with some equialization to see how bad the MeGUI file is.

[tebasuna51]

Seems a problem with your HT settings.
The Normalized 100% don't have any problem, only is 6 dB loud, but without clips.

Many users claim for more volume and you want less. Is dificult satisfy to everybody.

BTW, I never recommend use the Normalize function, only when use a downmix or experiment low volume.
This track have -25 dB RMS and don't need amplify the volume.

Cheers

[wuemura]

Originally Posted by tebasuna51

Seems a problem with your HT settings.

I don't think so, unless the person that listen to the file has some hearing loss any one can notice that the audio is saturated and if you EQ that, it will distort.

Originally Posted by tebasuna51

The Normalized 100% don't have any problem, only is 6 dB loud, but without clips.

This come back to previous post, what is best for the audio file, a 6dB boost, reduction or nothing? For this source is clear that 6dB was too much.

Originally Posted by tebasuna51

Many users claim for more volume and you want less. Is dificult satisfy to everybody.

I've never asked for more or less volume, I've suggested that MeGUI could use and automatic way to decide what to do with the audio file, instead of, a fixed value.

Is impossible to only use a reference value like 100%, 90% etc. You are deciding this values by guess instead of what you really need. Again, depending on the source that you are working with you need to cut the volume to a certain level and not 'always' boost, like this example.

 Analyzing...

    Gain   |  Peak  | Scale | New Peak |Left DC|Right DC| Track
           |        |       |          |Offset | Offset |
 --------------------------------------------------------------
  -6.01 dB |  32517 |  0.50 |    16278 |    0  |     0  | AUDIO.wav

 Applying Gain of -6.01 dB to file: AUDIO.wav

 WaveGain Processing completed normally

MeGUI lack this function, the point is not to remove a function that 'many users' are happy with, but to allow other people to have a second option and choose the best for what one needs.

Wavegain is opensource, anyone can download the source and add support for multichannel audio if this is the case.

Originally Posted by tebasuna51

BTW, I never recommend use the Normalize function, only when use a downmix or experiment low volume.

How much is low volume or when too much is too much?
One can't decide without a tool to analyze and apply the propper values.

Originally Posted by tebasuna51

This track have -25 dB RMS and don't need amplify the volume.
Cheers

This track was an example, not all captures has this value since I capture from DTV, VHS, Hi8, DV, Betacam etc.

[tebasuna51]

- I'm not a MeGUI developper, only a user. Then I say only my opinion.

- Everybody can listen your samples. Normalize never boost, only amplify, all dynamic range, until the max volume without distort.

- I know ReplayGain/WaveGain longtime ago, and I don't agree with the criterion (RMS values) used to calculate the recommended gain. For me is not the same a ballade and a hard rock.

- And everybody is free to use the soft desired. For me the ReplayGain/WaveGain concept is not recommended for movie audio tracks.