[MPlayer-dev-eng] Re: divx2svcd-0.5.1

[Nico Sabbi]

Andriy N. Gritsenko wrote:

>    Hi, Jan Knutar!
>
>Sometime (on Monday, April 26 at 11:12) I've received something...
>  
>
>>Do you know how these duplicate ones are handled when using -of mpeg? 
>>Zero sized frames in mpeg as well or something fun and accidental like 
>>that? 
>>    
>>
>
>    MPEG muxer write these as zero sized frames so it's not a problem. I
>think it's a problem of your mpeg video stream. Your reports about PTS
>earlier than SCR seems too suspicios, that is a problem and it _will_
>produce A/V desync. Unfortunately, mencoder can't tell muxer what is the
>max bitrate in stream so muxer tune it itself by estimation from current
>bitrate. But when you have it very little from beginning and then it
>improves too much then muxer cannot tune it fast and gets out of sync. :(
>May be if you could upload a little sample then I could try to find what
>to do with that.
>
>  
>
>>So, this problem would go away, if mencoder would when getting a 0 sized 
>>frame as input, feed the last proper frame again to libavcodec, or is 
>>it not as simple as this?
>>    
>>
>
>    If you do that you'll just get another intra-coded frame (i.e. the
>keyframe). It may break sequence (IBBPBBP... for mpeg1/2) and give you
>some weird results.
>
>    With best wishes.
>    Andriy.
>
>_______________________________________________
>  
>

out of curiosity:  what draft did  you follow, 11722-1?
I can't reconduce some of your pes headers to 13818-1.
A couple of hints, hoping you will turn them into patches :)
a) when the audio codec is ac3 (format 0x2000) you could use 
stream_id=0xbd, so demux_mpeg can correctly recognize ac3
(although maybe it's not enough:  IIRC there should be the extra 4 
leading bytes 0x....80 in the pes payload)
b) it video is mpeg2 it would be nice to set  correctly the mpeg1_mpeg2 
bit, to make some programs like replex happy

Do you have any idea why TMPG often stops demuxing mpeg files generated 
by mencoder long before the end?