[FFmpeg-devel] Adding a new output device

[Philip Prindeville]

> On Feb 28, 2018, at 4:12 PM, Nicolas George <george at nsup.org> wrote:
> 
> Philip Prindeville (2018-02-28):
>> I’m working on network test and measurement and I’ve been tasked to
>> write a test suite for streaming video performance.
>> 
>> This will run inside headless embedded devices (i.e. routers, probes,
>> etc) so has no hardware to render to.  It’s the streaming itself and
>> the end-to-end network transport that I’m interested in, observing
>> jitter, stalls, etc.
>> 
>> To that end, I wanted to write an output device that I can select as
>> the output driver to which the stream will be parsed and synchronized
>> to the appropriate frame rate, collect statistics for CBR/VBR/ABR,
>> etc. but which doesn’t actually do any rendering.
>> 
>> Whatever I write I’ll upstream, but it seems like it might be
>> generically useful.
>> 
>> I’m a network head and I’ve hacked DVB-S, so I’m vaguely familiar with
>> MPEG-4 AVC stream format.  I want to inspect enough of the stream to
>> be able to ensure synchronization with the frame rate, correctness of
>> the stream, etc. but not the expensive/unnecessary parts.
>> 
>> Any guidance for someone jumping into this for the first time?
> 
> Quick answer: look if "ffmpeg -i ... -f framecrc -" could be a good base
> for your project; and look at -f null for the wrapped-frame
> optimization. If you work at the muxer level, you do not need to know
> anything about the bitstream format.
> 
> Regards,
> 
> -- 
>  Nicolas George


Thanks, I’ll look at it shortly.

Will that level (muxer) give me visibility into synchronization so I can pace playback to 1:1 rate?

-Philip