There seem to be a lot of latency (many frames) when using the nv3dsink to display h.264 coming over UDP. It is an realtime application. Smoothness is not an issue, but latency is a big problem. I want the incoming packets to render at once. As soon as a frame is complete it should be displayed. My current gst pipe:
gst-launch-1.0 udpsrc port 15002 ! tsdemux ! tee name=t ! h264parse ! omxh264dec ! nv3dsink window-x=0 window-y=0 window-width=1920 window-height=1080 sync=false t. ! queue ! multifilesink
- How can it be improved in term of latency (optimally remove all buffering but the current incomplete frame)?
- Where are the frames getting buffered in this pipe?
- Is there a way to inspect the pipe while running to discover where buffering is taking place?