[FFmpeg-devel] MP4A-LATM support?

Luca Abeni lucabe72
Tue Nov 24 16:58:37 CET 2009

Hi Michael,

Michael Niedermayer wrote:
>> 1) LATM can be seen as a different bitstream syntax for AAC (I hope I
>>    am not oversimplifying anything, here... :)
>> 2) codecs in libavcodec will see the "regular" AAC bitstream (with codec
>> 3) muxers and demuxers in libavformat will see the LATM bitstream
>> 4) someone between codec and muxer is in charge of transforming the
>>    bitstream in LATM (this can be a bitstream filter, I suppose)
>> 5) how can the muxer know if the bitstream is "regular AAC" or LATM?
>>    I think we need a "fake" codec ID (CODEC_ID_LATM, or something like
>>    this) for identifying LATM streams. When the AAC stream is transformed
>>    in LATM (or vice-versa) by the bitstream filter, the codec ID is changed
>>    from CODEC_ID_AAC to CODEC_ID_LATM (or vice-versa).
>> Is this correct/reasonable? I am starting to work on the RTP muxer assuming
>> that some filter provides it with the LATM stream (and a LATM codec ID),
>> which simplifies things a lot :).
> did you read the thread with subj
> "Re: [FFmpeg-devel] [PATCH] seperate parser from codec selection"
> ?
> i think that also contains some related discussions

Yes, I read it (that's where I got the idea about having two different codec
IDs for AAC and LATM - the "fake" attribute was my own addition :), but I
missed the subthread about chained muxers/demuxers,

I now understand that the "someone" mentioned in my point 4) above should be
a muxer, not a bitstream filter.

I'll try some experiments with this chained muxers idea, thanks for pointing
me to this thread.


More information about the ffmpeg-devel mailing list