[vorbis-dev] packet rearranging questions

[Michael Smith]

On Wednesday 29 October 2003 17:55, Joshua Haberman wrote:
> Thanks for the reply!
>
> On Tue, 2003-10-28 at 16:27, Michael Smith wrote:
> > On Tuesday 28 October 2003 18:49, Joshua Haberman wrote:
> > > 1.  how exactly will the overlap/add nature of Vorbis packets affect me
> > > as I attempt to rearrange them?   Is it correct to say that for any
> > > Vorbis packet I take from an existing stream (other than the first
> > > packet) I must also take the packet preceding it?
> >
> > Yes. The reason that the first packet is an exception here is that it
> > produces no output audio (because the overlap is _always_ required).
>
> Ahh, I see.
>
> > > 3.  I notice that you can get decoders to discard some of the samples
> > > from the end of a packet by subtracting from the granulepos the number
> > > of samples you want discarded.  A similar way to do this for discarding
> > > samples from the beginning of a packet (used by vcut) is to set the
> > > granulepos to -x, where x is the number of samples you want discarded
> > > from the beginning.  But will this work in the middle of a stream,
> > > where the granulepos should reflect the total sample offset of the
> > > stream?
> >
> > No, a negative granulepos (other than -1) is undefined.
>
> Quoting from http://www.xiph.org/ogg/vorbis/doc/vorbis-ogg.html:
>
> "The granule (PCM) position of the first page need not indicate that the
> stream started at position zero. Although the granule position belongs
> to the last completed packet on the page and a valid granule position
> must be positive, by inference it may indicate that the PCM position of
> the beginning of audio is positive or negative.

Right. Note "a valid granule position must be positive", here. Remember, the 
granulepos is of the last packet in the page - it essential says "after 
decoding everything in this page, the final sample output will be number X". 
So (using some made-up numbers, these can't actually occur due to window-size 
power-of-two limitations), if you decode 1000 samples, but granulepos says 
"800", this infers that the _start_ of the audio in this page is at -200. 

This negative value is what is referred to in the paragraph below, and 
indicates that the first 200 samples should be discarded.

>
> [...]
>
> "* A negative value indicates that output samples preceeding time zero
> should be discarded during decoding; this technique is used to allow
> sample-granularity editing of the stream start time of already-encoded
> Vorbis streams. The number of samples to be discarded must not exceed
> the overlap-add span of the first two audio packets."
>
> > This only works at the beginning and end of stream, never in the middle.
>
> Bummer.  I suppose this will force me to create lots of logical
> bitstreams concatenated inside a physical bitstream to accomplish my
> goals.
>

Yes. The unfortunate downside... The alternative being some complex procedure 
where you re-encode only those parts at the boundaries, and reassemble as a 
complete single bitstream. Lossy and complex, though.

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.