[Ffmpeg-devel] dts decoding broken?

[Uoti Urpala]

On Fri, 2007-02-16 at 21:24 +0000, M?ns Rullg?rd wrote:
> Uoti Urpala <uoti.urpala at pp1.inet.fi> writes:
> > Does anyone intend to restore support for the downmixing functionality?
> 
> There is nothing to restore.  It used to *always* do downmixing with
> no option to get all the channels.

I know it used to always do that, and adding the ability to get
6-channel output is a good thing.

>   How would an application request
> downmixing?

I'm not sure what the exact syntax should be, but adding some field for
that (if nothing more complicated) shouldn't be too hard. I also just
checked the libdts documentation and it talks about "default behavior,
which is to apply the full dynamic range compression as specified in the
DTS stream". I'm sure that someone would want an option to turn that off
in situations such as a home theater, and since it talks about "as
specified in the DTS stream" it's probably not possible to get the same
effect by any processing after the decoder.

> > Can anything sanely use the current 6-channel output?
> 
> Works fine here.  Pretty much any motherboard less than a few years
> old seems to have 6-channel or better sound onboard, and there are
> many suitable PCI sound cards available.
> 
> > Is the channel order even documented anywhere in FFmpeg (outside
> > libdts)?
> 
> The order is whatever ALSA expects.

OK maybe it works with ALSA using a 6-channel setup (have you verified
all the channels really are right?), but at least ALSA doesn't seem to
downmix 6-channel playback properly for headphones by default on my
machine.

> I changed the libdts wrapper to decode all channels because Reimar was
> whining.  Now you and Sebastian are whining instead.  It seems
> impossible to satisfy all of you.

I haven't personally needed DTS decoding now so I'm not really
"whining". However I think downmixing (at least) should be available,
and I think using the codec downmixing would be better than trying to
create suitable conversion coefficients in MPlayer. So yes it's probably
impossible to satisfy everyone with only a single decoding behavior -
some kind of configurability is needed.