[MEncoder-users] Dolby downmix

[Reimar Döffinger]

On 30 dec 2010, at 19:04, Rolf Ernst <rolf.ernst at silverlightning.org> wrote:

> On 12/30/2010 6:34 AM, Reimar Döffinger wrote:
>> <snip>
>> With all the FFmpeg discussion (yes, suggesting it is ok, but please try
>> to keep the discussion a bit separate, hopping between programs makes it
>> a pain to follow the discussion) I haven't managed to actually
>> understand what the goal is.
>> Do you just want downmixing to two channels? Just use -channels 2
>> instead of -channels 6.
>> If you want to have a very specific downmixing matrix (instead of
>> something that should produce a reasonable 2-channel representation
>> from all 6 channels), use -channels 6 and -af pan with the precise
>> matrix coefficients.
> 
> Reimar,
> 
> what I want to is not to simply discard center, lfe, and back speakers but downmix so I can decode at least the Surround via Dolby II from the resulting channels (if possible).
> 
> Carl's response seemed to suggest that specifying -channels 2 would simply discard instead of mixing in information from the remaining four channels. Is this not so?
> 
> Now that you mention it I checked the doc and there is indeed an -af pan. The Nero aac encoder claims it is doing a Dolby II surround mix and so is Handbrake. Now, I listen to the Handbrake output through my HTS amp on Dolby II decode settings and I don't quite buy it but there is something coming out of the rear speakers so I don't quite know how well this can be done at all.
> 
> So the question again: My input file has AC3 5.1 Dolby audio and I would like to create a compressed output file that downmixes this information to Dolby II (Dolby II carries info for five speakers and the decoder can somehow pull this out of a dual channel signal).

Not quite, since that's not possible. However they are encoded in a way that for a lot of content will allow to get approximations of the original channel data.
And MPlayer can't encode to that format properly, but I can see two options:
1) use -af delay=...,pan=... with the delay filter approximating the phase shift
2) find a LADSPA filter that does the conversion (haven't found one myself, but I think there should be one).
However, as I understand it you want to re-encode that audio, I think that's probably a really bad idea, this kind of downmixing might severely break the psy models of the encoders, in which case re-encoding the full 6 channels is likely to give you better quality at the same bitrate...
>