Hello josh,

Great to hear you want to help us :)

I have been working on time stretching feature a while back, and I have
a branch[0] almost ready. The problem I got is that their are bugs
(unsupported features, FIXME in the source files) in 2 GStreamer
elements (GstVideoMixer for the video part, and GstAdder for the audio),
that we use in our standard timeline/pipeline, and make that the time
remapping doesn't work at all when using them. We found a hack to solve
this for the video part, and it looks to work properly, but the actual
solution would be to use GstVideoMixer2 instead of GstVideoMixer (in GES
we us the videomixer2 per default). And as for the audio, we should
implement the FIXME in the gst_adder_query_duration so our use case is
covered here.

Also all that is only for per clip video speed rate control. We can't
actualy do keyframe based time remapping yet because we are missing a
GstElement that would allow us to do it.

This subject is discussed here [1].

So I am not sure it is the simplest task you might want to start with,
because it is actually Gsteamer work, and I think it is quite
complicated things. If you think you have the competences to do it go
ahead, it would be great :), else you can have a look at the pitivi
love[2] page with a list of features (orders by difficulty of
implementation) we want to see implemented. Also you definitely should
come around #pitivi on freenode so we can help you get started ;)

Cheers, 

Thibault Saunier


[0] https://github.com/thiblahute/Pitivi/tree/timestreching
[1] https://bugzilla.gnome.org/show_bug.cgi?id=593828
[2] http://wiki.pitivi.org/wiki/PiTiVi_Love

On Sat, 2011-07-02 at 23:35 -0400, Benjamin M. Schwartz wrote:
> On 07/02/2011 12:04 PM, Josh Wang wrote:
> > My name is Josh Wang. I've just stumbled upon the PiTiVi GSoC website
> > and is particularly interested in the idea about "time stretching".
> > Although the application is over, to my knowledge this idea isn't
> > taken this year. Hence, I still want to contribute to this idea.
> 
> Great!
> 
> > I have a question, does this time stretching involves inserting frames
> > from in between frames?
> 
> In a sense, yes.  However, solving the temporal interpolation problem is
> not actually the limiting factor for "time stretching" in PiTiVi.  For the
> moment, we would be happy with the crude approach of simply displaying
> whichever input frame is closest to the desired time point!*
> 
> The problem we face is that PiTiVi's internal design does not yet have any
> (bug-free) way to express the concept of a video clip in the timeline
> running at variable speed ... and if you think about it you can see that
> this might be a nontrivial thing to integrate into the whole structure of
> a video editor, as it requires timelines within timelines.
> 
> If you're interested in the "software engineering" problem of designing,
> debugging, and building a user interface for time stretching in PiTiVi,
> your help would be most welcome.  On the other hand, if you're more
> interested in the "computer vision" problem of temporal video
> interpolation, your help would _also_ be welcome.  PiTiVi may not be ready
> to make use of a good temporal interpolator, but hopefully we will be soon
> ... and in the meantime, a good video interpolation library would be
> useful all across the Free Software universe.
> 
> Welcome!
> 
> --Ben
> 
> *: This could be regarded as temporal interpolation by convolution with a
> "box filter".
> 
> ------------------------------------------------------------------------------
> All of the data generated in your IT infrastructure is seriously valuable.
> Why? It contains a definitive record of application performance, security 
> threats, fraudulent activity, and more. Splunk takes this data and makes 
> sense of it. IT sense. And common sense.
> http://p.sf.net/sfu/splunk-d2d-c2
> _______________________________________________ Pitivi-pitivi mailing list 
> [email protected] 
> https://lists.sourceforge.net/lists/listinfo/pitivi-pitivi



------------------------------------------------------------------------------
All of the data generated in your IT infrastructure is seriously valuable.
Why? It contains a definitive record of application performance, security 
threats, fraudulent activity, and more. Splunk takes this data and makes 
sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-d2d-c2
_______________________________________________
Pitivi-pitivi mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/pitivi-pitivi

Reply via email to