Re: Key frame bitrate

You could send the patch to my email adress and tell me the base 
revision I have to use. I could do some experiments then. Is there a 
timeline for the 1.2 release? Is there a chance this feature gets into 
the (next) stable release?

> On Thu, Feb 10, 2011 at 4:29 AM, Andreas Engel<[email protected]>  wrote:
>> This sounds like a good idea. If you have a patch working for theora
>> 1.1.1 I would sure like to test it. But I need the possibility to
> I don't, I have a patch against the 1.2 (SVN) codebase.
>
>> connect to a running stream without getting garbage at the start. In the
>> current version I just have to wait for the first key frame. How can
>> this be done with "rolling intra" ?
> You would have to transmit an indication of how many frames ago the
> decoder must have started to decode in order to have fully recovered
> this by the point of this frame. The encoder could provide this
> information. Normally it would be a constant, analogous to a keyframe
> interval, but the encoder should have the freedom to vary it.
>
> For ogg files what we've discussed doing in the past is retroactively
> redefining the definition of the theora granpos. Right now it's
> defined so that the first part gives the number of the prior keyframe
> and the second part gives the number of subsequent frames since.  It
> can be redefined so that the first part gives the number of the frame
> where you'd have to start decoding and the second part is the number
> of frames since, and files like this just work in some player software
> that always do perfect reconstruction seeks (like firefox). They'll
> see to the target frame, read the "keyframe number", then seek to that
> and decode forward.
>