[gst-devel] Lip Sync with v4l2/oss/theora/vorbis/ogg

[Dave Baker]

Hi,

I've been trying for for a while now to get flumotion to capture audio and video, encode them to vorbis and theora and stream them over HTTP. The card I'm using doesn't support v4l1, so I can't use the standard wizard. I've tried making a v4l2 source component, and I'm now using a pipeline feeding a streamer (http://ystv.york.ac.uk/~dave/flutest4.xml), but they all have the same problem - the video eventually ends up leading the audio.

It seems to drift out at varying rates, but generally at about a second in half an hour or so. It also seems that can control is to some extent using the queue max-size-time and oggmux max-delay/max-page-delay settings, but setting them very low doesn't stop the sync drfting out by far more than the length of the buffers.

I asked about this on IRC and it was said that there were some sync patches applied to the v4l elements that weren't applied to the v4l2 elements. I can't test with a v4l1 source since I don't have any suitable hardware, but I'm starting to think this could be the problem. Does anyone have the details of this patch, or know a bit more about it? I'd be willing to have a go at applying the patch to the v4l2 code myself if I knew what it was.

If anyone can offer me a glimmer of hope, I'd be very grateful!

Thanks,


Dave