[FFmpeg-devel] Cooperative coding between (N)LE and FFmpeg; possibly quality improvements?

[Reimar Döffinger]

On Sat, May 07, 2011 at 09:21:13PM +0200, Stefan de Konink wrote:
> I have been thinking about implementing some sort of cooperative coding
> mechanism between a (non) linear editors and FFmpeg. For all current
> video's ffmpeg is a blackbox, it guesses what is a scene switch for example.
> 
> What if there was some sort of a 'hint' presented by an editor; make the
> next frame an intra-frame.

AVCodecContext.pict_type = AV_PICTURE_TYPE_I

> Or using the knowledge of spatial positions
> layers in the editor, or their movement to express prediction.

What is the purpose?
Just because the real movement is in some way doesn't mean
it is most efficient to encode it that way, nor does a scene
change mean it is necessarily a good idea to encode as an I-frame.
It seems unlikely to me that this will be useful for much besides
speeding up very low-quality (for the bitrate at least) encoding.