[Bug 776983] opusenc does not discard silence if DTX enabled

[GStreamer (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=776983

--- Comment #6 from Nubosch <andreas.kling at gmail.com> ---
many thanks for your review.

the buffer wrapped around the data in the adapter does not carry the time
information. Should this be generated by gst_audio_encoder_push_buffers?

Maybe it is a good idea to go the way speexenc did?
Just marking the buffer with GST_BUFFER_FLAG_GAP, and dropping it in
rtpopuspay?

Another option would be to add the event to gst_audio_encoder_finish_frame
where all required information seems to be available, but I'm unsure about
this. 

Sorry for bothering you with another question.
Do you think the dtx-drop feature should be optional?
Without a property this may break existing implementations.

The comfort noise generation in opus is not triggered,
also the jitterbuffer goes crazy, maybe dropping valid audio data.
This is the same situation for speex, but it may be acceptable depending on the
use-case.

having a dtx aware jitterbuffer would be cool, but this is another story, and
much more complicated than this simple changes...

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.