[FFmpeg-user] Adaptive Streaming between heaac and lcaac

Moritz Barsnick barsnick at gmx.net
Mon Jun 25 17:21:45 EEST 2018

On Mon, Jun 25, 2018 at 08:37:05 -0400, Ronak wrote:
> >>> HEAACv2 44khz/32kbps
> >>> HEAACv2 44/64
> >>> LCAAC 44/128
> >>> LCAAC 44/256
> >>> LCAAC 44/320
> I ran Apple's HLS mediafilesegmenter tool, and I can confirm that they set 1024 for the defaultSampleDurations in all moof segments for HE-AACv2 mp4s.
> Can this be fixed in ffmpeg? I'm happy to send a patch to the devel mailing list to correct this.

I assume you're using libfdk_aac for these? Or MacOSX's AudioToolBox?

It's really much more informative if you show us an actual command line
and the command's complete, uncut console output.

That said, I tried to understand fdk_aac's code, and it does seem it
makes some differentiation between LC and HE there - whether or not
correct I cannot tell. I believe it is *not* ffmpeg doing this.

> Another interesting thing I saw is that their generated fMP4s don't
> include sidx boxes, since the stream is audio only created from an
> AAC file. Why does ffmpeg include sidx boxes? Can those be removed?

I don't know much about MOV/MP4/fMP4 and atoms. I couldn't find
anything forbidding sidx though. If you think this is incorrect, or
inconvenient, please wait for other answers from fMP$ experts, or file
a bug ticket (if incorrect) or a feature ticket (if inconvenient, i.e.
you wish a flag to disable it), on https://trac.ffmpeg.org/

It would be good if you add a reference to some specification
describing the requirements more clearly.


More information about the ffmpeg-user mailing list