[FFmpeg-user] Video transcoding of the streamed data.

[Anil J]

Hi,

In HTTP streaming case (e.g. watching youtube), in my use case, I want to
adapt the video stream to suite my client requirement (e.g. codec
conversion). I am using a proxy server to tap the HTTP response messages
and in position to buffer the chunked video data.

I am however want to understand how I can send this data to FFmpeg. I want
to understand below:

1. How much of data (HTTP based video chunks) I need to buffer before I can
invoke the FFmpeg?
2. I am not sure how the content server (youtube) chunks the video? Are
these chunks logically independent of each other i.e. sliced at each
I-Frame? OR they are just sliced to fit the IP packet size without getting
into the video stream semantics.
3. In the case if they are just binary slices, how does FFmpeg would react
to a data which may not be complete. Note that, even if I buffer the
certain amount of chunks, I am not sure the validity of it, and whether
FFmpeg will like it.

Thanks,
/anil.