Composing Dolby Surround track from AC3 and mono WAV

[Evaldas]

Hi,

I have two files: english 5.1 AC3 track and a mono WAV in other language (translation of the original AC3, speech only). So I want to make one proper Dolby Surround WAV (which I can use later for SVCD or DivX) in translated language. I think I should take 4.1 channels (except center) from AC3 and mix rezulted WAV with translated WAV.
The question is, how to do it. I have some thoughts, I need to know am I doing this right.
Azid part I think should look like this:

azid -s surround -c normal -g xxdb -L 3db -d3/0 -ol,r

Next step is to mix obtained WAV with translated WAV. I think I will do it with CoolEdit Pro. I don't have many experience with such sound editing programs, maybe there is some important details, that I need to know. And, will the final WAV still have valid Dolby Surround information?

[sibe]

I think your azid parameters would not work for your task. -s surround and -ol,r are obsolete since these are the default values. -d3/0 would only give you the front speakers (left, center, right) from the ac3 track.
I think what you want to do is totally impossible with azid, because if you first downmix to a surround encoded file, afterwards mixing it with another mono track would destroy the surround information.

It may be possible with SoftEncode by replacing the original center track with the translated speech track. But this would only work right if the voices in the movie are in the center track only. It would be quite annoying to here the voice in one language from the front and in another language from the side.

Sibe

[Evaldas]

I don't think that "-d3/0" will give me only front speakers from AC3. Cause -d switch is downmixing mode, not channel selection mode (by default we all use "-d2/0", it gives us all channels downmixed to 2 scannels, not only the left and right form AC3). I didn't find in azid help how exactly this dowmixing mode works (I think I have to look at ATSC specification), but I think that with "-d3/0" the left channel would give me mixed left anf left surround channels, and the same with the right channel, and center channel will be the same as in AC3.
Since Sonic Foundry don't sell SoftEncode anymore, maybe its functionality is incorporated in other product by SF? Or they decided not support AC3 anymore.

[Taric25]

The first thing you need to do is convert your 5.1 track to 6 mono WAVs. Then, listen to the different channels. If you hear the voice of the actors in more than one channel, then you cannot make a surround conversion this way because more than one voice will be playing at the same time.

Even if all the voices are in the center speaker, simply replacing that with the mono WAV would destroy the surround information.

What I wonder is how an upmix is preformed on a 1.0 track. I was watching Disney's Snow White and the Seven Dwarves on DVD and afterwords I watched the special features. A sound person that works for Disney said that he alswys tries to preserve the original intent of the sound and always thinks "Is this what they were trying to do?" when he is doing a transfer to 5.1.

But how does he do this? How can you get 5.1 channels from a mono track? If we could figure this out, I think your problem would be solved.

[DarkAvenger]

Don't listen to what is said here. Most of them don't seem to know too much about matrix mixing.

"Easy way": Decode both tracks using HeadAC3he, but for the 5.1 deselect the center channel.

Most probably it is not advised to use two pass modes in this task, as otherwise you will get overflows at later stage. (Use a low global gain value, I can't help, which to use, just try and see.)

Now use CoolEdit and mix the alternative center to both stereo (DS encoded) tracks. The only problem is to find out the optimal gain value. I guess it will be something about 0.6 for the center.

The best to find out: Decode a part of the 5.1 with full channels, then the same part with no center and the same part of the 1.0. Now mix center into the file (both tracks use same phase) and try to find out the gain parameter for the center.

It is quite hard, I guess to get the optimal parameters, and working in 16bit space it not advised. You should at least make 24bit waves and once you have done the final mix, calc it down to 16bit if necessary.

Of course you can go the 6ch way, but then you have to manually mix down or try to make use of DSEnc, but that is not that easy...

[DarkAvenger]

I have an easier idea: *If* the center channel's volume is as high as the mono ac3's, simple decode both (5.1 w/o center into 2ch downmix) with the same global gain value and then mix the mono onto into both traks with 0.707. The only thing you have to make sure is that you don't get clipping.

[Taric25]

Originally posted by DarkAvenger
Don't listen to what is said here. Most of them don't seem to know too much about matrix mixing.

You're right. I've never even heard of matrix mixing before. Do you care to digress?

[DarkAvenger]

Uhh, what means "digress"? If you want explanations, go to my site and dl DSEnc and read the readme. It explains how DS encoding (=matrixed mixing, or dunno how it is precisely called) is done. Then if you know a bit of maths (highschool knowledge is enough), you'll find the correct conclusions.

[Taric25]

Digress means to explain in detail.

I read the readme. By the way, DSEnc on your site is DESEnc12.rar.zip so I think DESEnc12.rar is what you might want to fix that to. Anyway, I understand how a downmix works, but I still don't understand how one preforms a Dolby 5.1 upmix on a mono WAV. How did they do it for Snow White and the Seven Dwarves?

[DarkAvenger]

Well, I have not talking about an upmix, this topic is not about that. For that I cannpot really help.

BTW, I have my reasons, why I named it .rar.zip ...

[Evaldas]

@DarkAvenger, I've been away awhile and it seems you are the only who can help me.
Actualy that mono WAV is MP3 file (some volunteer translated the whole movie sitting with mic) which I decoded with MAD to 32bits per sample. And in fact it doesnt have any kind of posprocessing, normalization and etc. So I need to normalize it to some db value. How determine the value?

Originally posted by DarkAvenger
I have an easier idea: *If* the center channel's volume is as high as the mono ac3's, simple decode both (5.1 w/o center into 2ch downmix) with the same global gain value and then mix the mono onto into both traks with 0.707. The only thing you have to make sure is that you don't get clipping.

How to decode 5.1 w/o center into 2ch downmix? Will this azid line do this?
azid -s surround -c normal -g xxdb -L 3db -d3/0 -ol,r

[DarkAvenger]

a) Well then you gotta do trial and error
b) Use HeadAC3he. It uses azid engine. (I dunno much about azid cli.)