[FFmpeg-devel] [PATCH] als in mp4

[Alex Converse]

On Mon, Mar 23, 2009 at 2:51 AM, Jai Menon <jmenon86 at gmail.com> wrote:
> On 3/23/09, Alex Converse <alex.converse at gmail.com> wrote:
>> On Mon, Mar 23, 2009 at 2:29 AM, Baptiste Coudurier
>>
>> <baptiste.coudurier at gmail.com> wrote:
>> ?> Jai Menon wrote:
>> ?>> On 3/23/09, Baptiste Coudurier <baptiste.coudurier at gmail.com> wrote:
>> ?>>> Jai Menon wrote:
>> ?>>> ?>> [...]
>> ?>>>
>> ?>>> ?>>
>> ?>>> ?>> When were they initially set to the correct values?
>> ?>>> ?>
>> ?>>> ?> stsd atom/box? i think
>> ?>>> ?>
>> ?>>>
>> ?>>>
>> ?>>> Hummm, stsd does not contain the correct sample number for everything
>> ?>>> ?using 'mp4a' tag. Values are correct in 'esds' box.
>> ?>>
>> ?>> Which means we should parse the extradata (in this case als specific
>> ?>> conf) and get values from there? I can do this for als by maybe adding
>> ?>> a new function to mpeg4audio.c.
>> ?>
>> ?> I don't know how exactly ALS works, but it might use the
>> ?> AudioSpecificConfig like others, this code might belong to mpeg4audio.c,
>> ?> yes.
>>
>>
>> It's set in the AOTSpecificConfig. In this case it is the
>> ?ALSSpecificConfig for AAC it is the GASpecificConfig. Both of these
>> ?are embedded in the AudioSpecificConfig.
>>
>> ?Right now for AAC we just leave it set to zero and the decoder figures
>> ?it out just fine.
>
> That is fine by me too, the decoder can do this at init. But i doubt
> whether streamcopy will work correctly in the absence of a decoder.
> If it is too much of a hack, we can just add the codec id to the aot
> table and and leave it at that. ffmpeg -i will report wrong channel
> count though.
>
> [...]
>

I'm not necessarily arguing that we should do it that way. Just that
we should handle this situation the same regardless of codec.

--Alex