On Monday, July 25, 2011, כרמל מגן <[email protected]> wrote: > I have investigated the matter some more, and it seems that the delay is exatcly 2 frames (its 80 ms when I use PAL 25 fps video and 66 ms when I use NTSC 30 fps video). Furthermore, it appears when sending frames 1,2,3 av_read_frame does not return with frame 1 before I start sending frame 3. Does anybody have an idea why this might happen and how I can prevent it? > Thanks > Carmel
The stream has a setting of max b-frames. If you can control your source, simply set th his parameter to 0. Otherwise, you can work around the delay produced by libav, but when the b-frame actually arrives, you will have to wait, anyway. The behavior of libav is tuned to provide smooth playback with constant delay. Regards, Alex > 2011/7/19 כרמל מגן <[email protected]> >> >> Hello >> I have been integrating mpeg2 transport stream support into an application I'm developing using libavformat. >> After successfully transmitting my video (h.264 frames encoded by the calling application) and receiving it, I did some latency testing. >> To my surprise, I calculated 80 ms between the time that I finished sending the frame in one process (av_write_frame returns) until I finish receiving it in the second process (av_read_frame returns and the packet contains the same frame). >> To identify the frame I used the pts field of the packet. (i.e I printed the time when sending together with the pts of the frame, and then the time of receiving together with the pts of the frame, and then calculated the difference). >> This was tested using multicast transmission between 2 processes on the same computer (on two different machines sitting on different networks) so I don't believe network latency can account for it. >> Can anybody explain what might be taking all that time and how I can reduce it? My application demands very low latency. >> Thanks >> Carmel
_______________________________________________ Libav-user mailing list [email protected] http://ffmpeg.org/mailman/listinfo/libav-user
