Mixing lossy live sources

Nicolas Dufresne nicolas at ndufresne.ca
Thu May 17 15:21:28 UTC 2018


Le jeudi 17 mai 2018 à 12:52 +0100, Tim Müller a écrit :
> On Thu, 2018-05-17 at 12:28 +0200, Benjamin Le Bozec wrote:
> 
> Hi Benjamin,
> 
> > I'm mixing two RTMP live sources with the compositor element : 
> > rtmpsrc (do-imestamp=true) ==>
> >                                                        compositor
> > ==>
> > rtmpsink (sync=true)
> > rtmpsrc (do-imestamp=true)  ==>
> > 
> > But the sources are coming from cellphones, through a lossy network
> > (the cellular network), and packets are lost from time to time.
> > Consequently, the compositor elements tends to freeze, waiting for
> > both streams to deliver packets. 
> > 
> > Is there a way to ensure a continuous streaming, that is for the
> > compositor to continue to deliver images even if some (or all)
> > sources do not deliver images ?
> 
> compositor *should* be able to handle this scenario well. However, it
> will do so only if the source elements have identified themselves to
> be
> 'live' sources (like rtspsrc/udpsrc). I don't think rtmpsrc does
> that.
> 
> You *might* be able to hack around it by using
> gst_base_src_set_live()
> and setting the "latency" property on compositor to some reasonable
> value; or add a pad probe and intercept the latency query from
> compositor to the sources.

I believe you can turn this into a lice source using the following
construction, let me know if that works for you.

  rmptsrc do-timestamp=1 ! identity sync=1 ! ...

You can place the identity elsewhere if needed.

> 
> Not sure if rtmpsrc will gracefully handle this case though (drop
> packets to 'catch up').
> 
> Cheers
>  -Tim
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 195 bytes
Desc: This is a digitally signed message part
URL: <https://lists.freedesktop.org/archives/gstreamer-devel/attachments/20180517/c9ea56e7/attachment.sig>


More information about the gstreamer-devel mailing list