[MPlayer-dev-eng] AAC decoding is out of sync

compn tempn at twmi.rr.com
Fri Apr 16 14:09:33 CEST 2010


On Fri, 16 Apr 2010 12:11:54 +0200, Grozdan wrote:
>On Fri, Apr 16, 2010 at 6:58 AM, RC <cooleyr at gmail.com> wrote:
>> On Thu, 15 Apr 2010 18:57:32 +0200
>> Grozdan <neutrino8 at gmail.com> wrote:
>>
>>> I think it was a bad call to add ffaac as default decoder for AAC as
>>> it tends to screw up in a lot of cases. I prefer FAAD over it anytime.
>>
>> The MPlayer project is perpetually in "testing".  There is no stable
>> branch.  Faster codecs have always gotten precidence over more reliable
>> ones, and setting somewhat buggy codecs as the default has long been the
>> method to get them tested.  In short, it's not an argument you can win.
>
>I'm not looking to win arguments here. Just pointed out that IMO it
>was a bit too soon that ffaac was made default and from a user's
>perspective who just wants to throw a file at it and watch it, he
>doesn't really care which one is default as long at it works. There
>are much more users who are not interested in testing than there are
>who want to test or trying for hours to find a solution to get it work
>(and get frustrated along the way, some eventually giving up). And
>last time I checked, peeps involved in mplayer & ffmpeg already do a
>good amount of testing.

we got 4+ bugreports out of that simple change already. its only done
in svn, not an actual release. and that user wanted to watch matrix 3
anyways. i'd consider it a feature not to hear the dialog in that movie.

but yes, ordinary svn users were inconvenienced for a month while the
change was in effect. this is how mplayer and ffmpeg have worked
since the projects started.

mplayer/ffmpeg devs test all files we can get a hold of. this was a
prerequisite of committing the aac decoder in the first place. we
force users to test files because there is no way we can check every
file ever made ourselves.

an automated testing script (like ffmpeg's fate) would be very useful to
mplayer in checking for regressions. but mplayer has a very small
amount of developers, so i dont think it will get done without outside
help.

>Only yesterday we had a pretty frustrated user come to #mplayer asking
>why a52drc no longer works as he wasn't aware that mplayer has

thats a good point. maybe there should be some kind of message like
'warning, a52drc works with afm liba52 but decoder is afm ffmpeg' or we
should make a52drc work with the ffac3.

>switched to ffac3 which although has a drc option, it doesn't get
>exposed to mplayer like the one from liba52

there is a patch here for ffac3 drc:
http://bugzilla.mplayerhq.hu/show_bug.cgi?id=1607

how come no one suggested he use -ac faad or -af pan to mix the 6
channels to 2? i suggested -af hrtf and user reported that it helped,
but caused some buzzing noise (an hrtf bug? needs to be reported.)

i know you just want to throw a file at it without messing with command
lines, but how do you think we got so many files working without extra
options in the first place?

ac3 dialog is known to be at lower volume than all other codecs.
http://roundup.ffmpeg.org/issue1153
it happens with faad too. faad does downmixing and you dont notice it
as much.

-compn



More information about the MPlayer-dev-eng mailing list