[gst-1.0] Make videomixer mix raw frames without timing infromation

[Sebastian Dröge]

On Mi, 2015-03-25 at 03:28 -0700, Arjen Veenhuizen wrote:
> Hi Sebastian,
> 
> Thank you for your swift reply. I wasnt aware of the compositor element,
> will give that a try. 
> 
> Basically, this is what I want to do (very simplified pipeline):
> 
> 
> fifo0.pipe and fifo1.pipe are getting stuffed with raw yuv frames in two
> different threads by two different pipelines at exactly 25 fps (using a
> custom filter).  (The filesrc thus functions as a "live" source) This custom
> filter also takes care of throwing away frames if its internal queue fills
> up, making sure that the lag between input and output is always just a
> couple frames at most.
> 
> As the frames arrive at the compositor, the compositor requires them to be
> timestamped. If timestamping is not identical, the compositor will wait for
> frames to arrive on its sinks with matching timestamps. I suspect that my
> timestamping is off between different compositor sink branches (because that
> requires out-of-band communication between the different sink branches),
> causing unwanted delays. Best for me would be if the compositor element just
> mixes frames as they arrive.
> 
> Any more hints?

Timestamp the frames with the current running time from the clock on
arrival in the pipe, i.e. what do-timestamp=true would do. And then use
compositor from GIT master (not 1.4). It should do the right thing then,
even if you drop some frames in your custom filter.

> In the mean time I will check-out the merrits of the
> compositor element. Am I assuming correctly that the compositor element
> superseeds the videomixer element?

Yes

-- 
Sebastian Dröge, Centricular Ltd · http://www.centricular.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 949 bytes
Desc: This is a digitally signed message part
URL: <http://lists.freedesktop.org/archives/gstreamer-devel/attachments/20150325/aaac964a/attachment.sig>