[Qt-interest] jerkiness in video on using phonon on windows

[Kunal Lagwankar]

hi Jason,

We have resolved the issue. The video decoder was sending 3 frames (SPS, 
PPS and I frame) just before the I frame. So the clock was expecting 3 
timestamps thats why it was jumping ahead and coming back for the first 
P frame.
To resolve this, we returned false in case of SPS and PPS which then 
does not forward the frame (in this case, zero packet) to the video render.
However, this does not explain why it was playing properly (and still 
plays with the above condition) in graphedit.

Thanks for your suggestions.

Regards,

Kunal

Jason H wrote:
> Some questions/tests
>
> Is the QIODevice sequential? (socket?)
> With 2-3 second jerkiness, it sounds like the codec (or i/o) is going back to the last I frame after exhausting P frames. Can you transcode it to be all I frames?
> I think it may  be a problem with a partial read (in that you don't have a complete I/P frame) and that the Codec is resuming from the last I-frame. Meanign to me, it could be a buffering problem. Which then begs the question, if the media in a QBuffer? Does it work when it is?
>
>
>
>
>
> ----- Original Message ----
> From: Kunal Lagwankar <kunal at novix.in>
> To: qt-interest at trolltech.com
> Sent: Fri, November 20, 2009 9:55:25 AM
> Subject: [Qt-interest] jerkiness in video on using phonon on windows
>
> Hi,
>
> We are subclassing QIODevice to pass audio-video data to a custom parser 
> (a Directshow filter implementation) using Phonon. Phonon creates a 
> graph that includes the custom parser and corresponding audio-video 
> decoders and rendering filters. The custom parser internally separates 
> audio and video, which is then decoded and rendered.
> When this sequence is played, jerkiness is observed periodically (spaced 
> every 2-3 seconds). By video jerkiness, I mean that previous frame is 
> again rendered causing a break in continuous viewing. It is important to 
> note that there is no break in audio.
>
> When the same file with the custom format is drag-n-dropped in to 
> Graphedit, it automatically creates a pipeline and renders audio-video 
> data satisfactorily (meaning no jerkiness and in audio-video 
> synchronization). From this, we have concluded that the directshow 
> filters by themselves do not have any bug and perform correctly.
>
> We are unable to figure out why this behaviour is observed on using 
> phonon in windows. Any help/suggestions in this matter will be appreciated.
>
> Thanks.
>
> Kunal
>
> _______________________________________________
> Qt-interest mailing list
> Qt-interest at trolltech.com
> http://lists.trolltech.com/mailman/listinfo/qt-interest
>
>
>
>       
>   
> ------------------------------------------------------------------------
>
>
> No virus found in this incoming message.
> Checked by AVG - www.avg.com 
> Version: 8.5.425 / Virus Database: 270.14.76/2517 - Release Date: 11/21/09 07:47:00
>
>