[FFmpeg-devel] FFmpeg booth at FOSDEM 2010

[Stefano Sabatini]

On date Monday 2010-02-08 20:30:33 +0100, Stefan Gehrer encoded:
> compn wrote:
>> On Mon, 08 Feb 2010 00:13:41 +0100, Vladimir Pantelic wrote:
>>> this year we had 6 "machines" running lavf/lavc at the same time:
>>>
>>> http://www.youtube.com/watch?v=9pwUdRKllo0
>>
>> wow thats a big ffmpeg poster! it looks larger than any other posters
>> i've seen in the fosdem vids.
>>
>> were there any interesting comments or questions for the ffmpeg devels?
>
> Here are a few comments I heard, well they were more like
> feature requests:
>
> - Someone who cross-compiles ffmpeg with just a hand-picked
> selection of codecs expressed his wish for an interface which
> simplifies setting configure options, similar like kconfig with
> a graphical or an ncurses interface.
>
> - Someone using ffmpeg in a PHP-based web application said
> he wants to display container information to the user and has
> a lot of work parsing ffmpeg's diagnostic output as it differs
> depending on format
>
> - The same person also has to go through some hoops as he
> aims at "automatic storyboard generation". He currently uses
> ffmpeg to export a video to MJPEG and feeds that into another
> program which does a scene cut detection. Then he creates an
> index file listing the frame positions in the middle between
> the detected cuts, exports all frames of the video with ffmpeg
> to image files and takes only the frames that are in the index
> file to have his storyboard as a result.
> Features simplifying this would be first to expose the
> scene-cut detection which is buried in some encoders to the user
> and secondly to accept a list of positions for doing screenshots
> in a batch.

All this may be done with a libavfilter video chain, there are plenty
cut-detection filters out there so picking one and adapting to lavfi
should be easy enough.

> Apparently he also has an experimental setup where he
> combines this with speech recognition, so that he can get
> a complete storyboard of pictures and textual dialogues out
> of a video file. That sounded quite interesting to me.

That may be done adding audio support in libavfilter and implementing
a speech recognition library wrapper filter (af_sphinx?).

Yes libavfilter is my hammer and everything looks like a nail to me
;-).

Regards.
-- 
FFmpeg = Foolish and Funny Monstrous Pure Ecumenical Game