[FFmpeg-devel] parser for the nellymoser decoder

[Benjamin Larsson]

Hi,

Michael Niedermayer wrote:
> Hi
> 
> On Sat, Oct 06, 2007 at 11:26:47PM +0200, Benjamin Larsson wrote:
>> Hi, do I need to code a parser for the nellymoser codec when the audio
>> tags only contain 1,2 or 4 64byte frames ? There is no header to find
>> the only thing to use is the block_size.
> 
> well, why are 1,2 or 4 frames concatenated?
> 
> is it because of some requirement f the container? or is it arbitrary
> and would any packing work if it where muxed?
> 
> would a (flv?) muxer work if it received only 4 frame packs? would it work
> if it receive only 1 frames?
> if these blocks ever need to be split than they definitly need to be split
> in the demuxer or parser
> 
> can this packing change within a file, that is like 1,2,4,4,4,2,1,1,1 ...
> or are there always the same number of frames packed together?
> if it can change then they must be split in the demuxer or parser as
> the alternative will make muxing in some containers much harder
> 
> if none of the above applied then its not strictly needed to spit but
> its IMHO still a good idea to split them ...
> 
> [...]
> 

I think it's the same case as swf_adpcm, there can't be to many audio
tags per time unit for the flv format when it's muxed with video. The
nellymoser format also has 3 different sample rates, 8,11,22 which could
map to 1,2 and 4 times 64 sizes. If that is the case then a mux to
another format and back would work without any trouble. If the blocks
where split then a muxer would need to concatenate the blocks depending
on the sample rate. So I actually think the right way to do it is not to
split the tags into single frames.

All in all this is just guesses as there is no spec.

MvH
Benjamin Larsson