<html> <head> <style></style></head> <body class='hmmessage'><div dir='ltr'>Thanks Nicolas. I have managed to fix the elements and when I get a minute will post a but with relevant patches. I ended up doing patches on 0.10 and 1.0 as I needed compatibility with gstreamer-java. However seeing how much things seem to have improved in 1.0 I may end up either using gstreamer1.0 directly or update gstreamer-java (though this looks like a few weeks of work potentially). In gstreamer1.0 my patch set the latency to GST_CLOCK_TIME_NONE - in gstreamer0.10 I just guessed as it failed an assert - really didn't make much difference to my scenario. For VP8 I set timebase to 90000 if there is no framerate (this is the clock-rate noted in the RTP payloading spec).<div><br></div><div>The pipelines I have seem to work pretty well with the patches I've made -I'm not really too concerned about dynamic latency as if I don't have frames coming in means I can't create any better frames going out. The pipelines seem to push through frames as quickly as they receive them as far as I can tell from limited testing. I am not sure what you mean by 'live' I think you mean camera, but I guess the same applies to using RTP streams (videoconferencing and video calling scenarios). I appreciate the idea of using videorate, but I think that may not really help as I want to just send frames from the encoder out ASAP.</div><div><br></div><div>I think I can see your changes on that bug, but I don't understand the effect of the latency on the video encoders. I may need your patch as well to cover some scenarios when RTP streaming. Perhaps you could explain your patch and sink recovery mode.</div><div><br></div><div>At least I don't seem totally out of the ball park and my patches stop gstreamer crashing.</div><div><br></div><div>Thanks</div><div><div><br><div>> Date: Thu, 07 Mar 2013 22:04:45 -0500<br>> From: Nicolas Dufresne <nicolas.dufresne@collabora.co.uk><br>> To: Discussion of the development of and with GStreamer<br>> <gstreamer-devel@lists.freedesktop.org><br>> Subject: Re: Best way to fix video encoder issues with 0/1 framerate<br>> Message-ID: <1362711885.5234.5.camel@nicolas-laptop><br>> Content-Type: text/plain; charset="utf-8"<br>> <br>> Le lundi 04 mars 2013 ? 14:27 +0000, Thomas Greenwood a ?crit :<br>> <br>> > I currently favour 3 and I'm not sure that I can really do the others,<br>> > plus it makes this easier and more robust for others without<br>> > introducing extra elements into the pipeline. Also I have made the<br>> > patch to x264enc work, so just need to do the equivalent on vp8enc.<br>> <br>> <br>> Note that having dynamic framerate also mean having dynamic latency. I<br>> have made an attempted [1] to implement dynamic latency in 0.10, but it<br>> might not be the right way. A forth solution would be to fix a framerate<br>> using videorate. This will not fix the latency issue as videorate does<br>> not work in live pipeline, but might work for your usecase. Making<br>> videorate work in live (timeout based) would also be really nice.<br>> <br>> Nicolas<br>> <br>> [1] https://bugzilla.gnome.org/show_bug.cgi?id=675761<br>> <br>> -------------- next part --------------<br>> An HTML attachment was scrubbed...<br>> URL: <http://lists.freedesktop.org/archives/gstreamer-devel/attachments/20130307/6821399f/attachment-0001.html><br>> <br>> ------------------------------<br></div></div></div> </div></body> </html>