[Ffmpeg-devel] h264/rtp transport (RFC 3984)

Ryan Martell rdm4
Thu Oct 5 00:00:36 CEST 2006

Hi Michael--

That was it;  everything is working now.

I'll work on cleaning it up and getting it together to submit next week.


On Oct 4, 2006, at 3:26 PM, Ryan Martell wrote:

> Hi Michael--
> Okay, I'll look for that as well.
> To make sure I understand:
> The decoder needs to have all the nals for one frame (based on  
> timestamp?) in the same AVPacket.
> Currently, I'm just using the same code in rtp.c, and modifying it,  
> so once I hand it a packet, I have no idea what it does with it  
> (I'll follow that path as well).
> If not with timestamp, how do I know when there is a new frame?
> Thanks!
> -Ryan
> On Oct 4, 2006, at 3:14 PM, Michael Niedermayer wrote:
>> Hi
>> On Wed, Oct 04, 2006 at 02:40:33PM -0500, Ryan Martell wrote:
>>> Okay, so I'm really, really close to implementing h264 over rtp (per
>>> RFC 3984).
>>> I am getting video, but there is lots of grey.  I am getting lots of
>>> "Concealing 150 MV errors, .." messages.
>>> I have looked over my byte manipulations all morning, and I am  
>>> pretty
>>> sure that I am decoding the stream correctly, and creating the AvCC
>>> stuff from the SDP announce (I have verified that my code produces
>>> the same avc block as the file that is sourced from the server).  I
>>> am prepending a 4 byte NAL packet length to the start of every
>>> packet, so it goes through the AVC loop just fine.  Byteswapping all
>>> looks good.
>>> So... I'm thinking that the problem has to do with my timestamps and
>>> sequence numbers.  I'm going to start looking into this right now,
>>> but if anyone knows whether that would account for lots of grey
>>> (including grey on the bottom half of the full frames?), I would
>>> appreciate knowing that I am on the right track.
>> this sounds a little like you feed incomplete frames into the  
>> decoders
>> this is not supported concatenate all packets for one frame and then
>> give it to the decoder, you can also use a AVParser to do that for  
>> you ...
>> [...]
>> -- 
>> Michael     GnuPG fingerprint:  
>> 9FF2128B147EF6730BADF133611EC787040B0FAB
>> In the past you could go to a library and read, borrow or copy any  
>> book
>> Today you'd get arrested for mere telling someone where the  
>> library is
>> _______________________________________________
>> ffmpeg-devel mailing list
>> ffmpeg-devel at mplayerhq.hu
>> http://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at mplayerhq.hu
> http://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel

More information about the ffmpeg-devel mailing list