[FFmpeg-devel] [FFmpeg-cvslog]?r10889?-?trunk/libavcodec/allcodecs.c

[Michael Niedermayer]

Hi

On Thu, Nov 01, 2007 at 02:30:08AM +0200, Ismail D?nmez wrote:
> Thursday 01 November 2007 Tarihinde 02:31:54 yazm??t?:
> > On Thu, Nov 01, 2007 at 02:17:45AM +0200, Ismail D?nmez wrote:
> > > Thursday 01 November 2007 Tarihinde 01:58:27 yazm??t?:
> > > > On Thu, Nov 01, 2007 at 01:33:20AM +0200, Ismail D?nmez wrote:
> > > > > Thursday 01 November 2007 Tarihinde 01:38:22 yazm??t?:
> > > > > [...]
> > > > >
> > > > > > > > Oh, and the best way to fix this anyway is to drop support for
> > > > > > > > libvorbis decoding ;-)
> > > > > > >
> > > > > > > Or remove non-working vorbis encoder.
> > > > > >
> > > > > > or remove the libvorbis encoder, that way the vorbis encoder will
> > > > > > soon be fixed by volunteers ...
> > > > >
> > > > > Would you like to bet on that? :)
> > > >
> > > > sure, would a soon = "5 years" be ok? :)
> > >
> > > Heheh, you got the point ;-)
> >
> > would you bet on some volunteer implementing it in 5 years if libvorbis
> > stays default?
> 
> No, but writing a good audio encoder is very hard afaik due to 
> psyco-acoustics.

no not at all, writing a good audio encoder is hard because noone starts
writing one not because any specific part would be, actually psyco-acoustics
are completely optional, so until we have a well written encoder without
psyco-acoustics the subject psyco-acoustics is limited to lazy trolls
who search for an excuse not to start working

and really, psyco-acoustics is AFAIK just running a FFT over the input and
from that calculating with a mayby 10 lines of code how audible quanization
errors are at various frequncies and then just selecting rate-distrortion
coeffs (these lambda thingys) for all bands after that its just normal non psy
encoding, that is find the encoding with minimum
rate + distortion[band]/lambda[band] for everything

i did suggest the above already for the vorbis encoder but was just ignored
...


also the reason why most people have a huge problem with psycho acoustics
seems to be IMHO that its not explicitly specified in the mp3 standard
how to do it (and of course thats the way it should be it doesnt belong
in the mp3 spec ...)
these people expect that they can write a perfect encoder by just
implementing, line per line what is in the spec

so really, please everyone either shut up or start writing code, iam sick
of these vague psycho acoustic is hard discussions from people knowing
about psycho acoustics as much as a butcher from making bread
go use google / citeseer and read 5 papers about psycho acoustics, if you
still think its hard afterwards then iam sure we would all benefit from
your (not you specifically but generally) comments but currently we all
do just random guessing about the difficulty of psycho acoustics while
we dont know _anything_ about it thats just a waste of time ...


[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

The greatest way to live with honor in this world is to be what we pretend
to be. -- Socrates
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20071101/5e5be552/attachment.pgp>