Speex dev - May 2009 - newbie qs. how to seek to a point in the Speex file?

Thanks for your reply.
I am using JSpeex so it is not possible to use the libs you mention.
I looked a little at the source in http://jspeex.sourceforge.net but cant see
where the length can be stored. Unfortunately, that project does not seem to be
maintained - no replies to queries posted on the forums.
Does anyone know how to store the length of a JSpeex clip in the header?
thanks,
Anil


--- On Wed, 4/22/09, Conrad Parker <conrad at metadecks.org> wrote:
> From: Conrad Parker <conrad at metadecks.org>
> Subject: Re: [Speex-dev] newbie qs. how to seek to a point in the Speex
file?
> To: goodnewsforyou at yahoo.com
> Cc: "speex dev" <speex-dev at xiph.org>
> Date: Wednesday, April 22, 2009, 6:29 AM
> 2009/4/22? <goodnewsforyou at yahoo.com>:
> > Although this actually pertains to JSpeex, I am asking
> these questions here
> > because they are general questions and also the JSpeex
> project does not seem
> > to be supported any more.
> > 1) Can the duration of the clip be stored in it? Page
> 24 here
> > http://www.speex.org/docs/manual/speex-manual.pdf
> describes the format, but
> > I could not see where the duration of the clip is
> stored.
> > 2) Since the Speex compression is not linear, is it
> still possible (say) for
> > a 60 second clip which is 60Kb, to seek accurately to
> 10 seconds on the
> > time-line by discarding the first 10Kb of data? If
> there is any example of
> > how to do an approximate seek, it would be
> appreciated.
> 
> Hi,
> 
> If you're talking about Speex data in an Ogg container
> (which is the
> format produced by speexenc) then:
> 
> 1) no the duration is not explicitly stored in the Ogg
> Speex file. You
> need to read the granulepos (timestamp, in units of audio
> samples) of
> the last Ogg page in the stream.
> 
> 2) to seek you need to seek for the Ogg page containing the
> data.
> 
> I suggest you use liboggz to handle the Ogg reading and
> seeking:
> http://www.xiph.org/oggz/. Info about the seek api is
> in
> http://www.xiph.org/oggz/doc/group__seek__api.html
> 
> cheers,
> 
> Conrad.
>