xxxFrameDesc objects enable us to copy parameters of input to output

[cchampet]

They should contain an AVCodecContext.
And next use avcodec_copy_context [http://ffmpeg.org/doxygen/trunk/group__lavc__core.html#ga3933b21b9dcb9173df3e56673b7a7d76] to set the encoder

[valnoel]

Why don't we analyse the input file to get the input properties, which would be overridden by the profile parameters, and then set as output properties ?

InputFile ==> Properties ==> OutputFile
                  ^
                  |
               Profile

[valnoel]

AudioFrameDesc must also contain the corresponding video fps. Maybe one fps value could be set to the whole OutputFile, and shared by all the streams (video, audio, etc..) ==> in FormatContext ?