<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"> <html> <head> </head> <body bgcolor="#ffffff" text="#000000">     Dear all, I am using a pipeline to do video streaming with the following structure:     V4LSRC -> H264_ENCODER -> MPEGTSMUXER -> UDPSINK If I insert a print message into the H264 encoder when some frame must be discarded because it is late then it happens this: 1.- Using a high QP (> 25), there is no problem (no frames are skipped at all). 2.- When I use a smaller QP (< 10), the encoder element starts skipping frames (and printing the message I added). But if I change the MPEGTS and the UDPSINK by a FAKESINK, no frames are skipped at all. So problem is not the video source and the encoder! But I have also checked that the MPEGTS muxer and the UDPsink give speed enough for the bitrate, so they are not the problem neither! Splitting the pipeline every element is doing well, but the whole pipeline does not work! Why? What is happening? It is clear that when you analize each part of the pipeline, it works perfectly, but when you connect all the elements they block each other or something like that, and some frames must be skipped. How can I optimize this? How can I do to have the video source plus the encoder working "independently" from the other part of the chain? How many threads are working there? (I have added a 'queue' element after the encoder, without improvement).     Thank you very much, any help is appreciated.     Thanks, LD. PS: The real pipeline is: gst-launch-0.10 v4lsrc capture-width=720 capture-height=576 fps-n=25 fps-d=1 ! hwencoder width=720 height=576 codec=h264 qp=20 iinterval=50 framerate=25 ! 'video/x-h264,width=720,height=576,framerate=25/1' ! mpegtsmux ! rtpmp2tpay ! udpsink host=192.168.0.5 port=5000 </body> </html>