On Wed, Apr 27, 2011 at 11:31 AM, anders musikka
<[email protected]> wrote:
> Hello!
>
> My name is Anders Musikka, and I'm new to this list.
>
> Firstly, let me give a big thank you to everyone contributing to MLT!
> It seems to be a very capable, well designed, and needed open source
> library.
>
> Recently, I was trying to use openshot to edit a few vacation videos.
> I had some trouble. Being a programmer and open source enthusiast, I
> traced one of the problems back to MLT.
>
> What had happened was this (I think):
>
> 1: Openshot had stored a slightly invalid length of a few of my clips.

It gets that info from Mlt. Mlt gets that from libavformat, but it is
only an estimate. Really, one needs to parse out and possibly decode
all frames to know the true length. The calculation with Mlt to
determine length from estimate changed recently to be more
conservative by a few frames as the often liberal estimate was causing
problems for some people who are using the whole length of their clip.

> 2: Openshot specified my clip as a producer for MLT, entering an "in"
> and "out" value, where the "out" value was actually one frame after
> the real last one in the clip.
> 3: Openshot then added a few transitions. However, since the actual
> length of the clips, as rendered by MLT, weren't the same as the
> nominal length (out-in+1), the transitions weren't placed correctly on
> the timeline by Openshot, giving unsightly flickering at the start of
> fade-ins, for example.
>
> Now, it could probably be argued that there's a bug with openshot.
>
> However, the fact that anyone who uses MLT to compose complex videos,
> needs to do a sort of dead reckoning, to try and figure out when in
> absolute time each clip will actually start, invites this kind of bug,
> and makes it very hard to debug when it occurs.
>
> Would it be possible to change MLT, so that the rendered length of a
> clip would always be exactly (out-in+1) frames? I.e, if there aren't

That is the intention, so there is probably a bug. I think this is
related to the bug report that caused me to make the length estimate
more conservative. Moreover, I think it was a regression as people
seemed to reporting problems related to this since mid-2010.

> enough frames in the clip, simply repeat the last frame as required?

Yes, it is possible. The current design is that it should produce what
it calls "test" frames - think of them as dummy frames, which are
white silence by default. But we can have an option to hold the last
frame from the producer.

> Maybe provide this behavior as a kind of option?
>
> I'm willing to do some implementation work for this, however I would
> need to be pointed in the right direction to be able to come up with a
> patch.

It is probably some particular interaction between the avformat
producer src/modules/avformat/producer_avformat.c. I will have to do a
lot of analysis to give any more detail. I will work on this.

> best regards,
> Anders Musikka

------------------------------------------------------------------------------
WhatsUp Gold - Download Free Network Management Software
The most intuitive, comprehensive, and cost-effective network 
management toolset available today.  Delivers lowest initial 
acquisition cost and overall TCO of any competing solution.
http://p.sf.net/sfu/whatsupgold-sd
_______________________________________________
Mlt-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mlt-devel

Reply via email to