Re: [Debconf-video] MiniDebConf Prishtina video
On Sun, Oct 08, 2017 at 10:11:46PM +0200, Daniel Pocock wrote:
> On 08/10/17 21:14, Wouter Verhelst wrote:
> > Hi Daniel,
> >
> > On Sun, Oct 08, 2017 at 08:09:23PM +0200, Daniel Pocock wrote:
> >> Hi,
> >>
> >> Where can I store the raw video and audio files from the MiniDebConf in
> >> Prishtina?
> >>
> >> Is there a workflow I can suggest for volunteers to process these into a
> >> format suitable for the web?
> >>
> >> We have about 30GB of data, including video files from a DSLR and audio
> >> recorded with a lapel microphone.
> > Whoa.
> >
> > I hope you used a clapperboard or something along those lines, otherwise
> > getting them back into sync might be complicated. The human mind is
> Given that over half the talks are in Albanian and they are all
> concatenated in 3 large WAV files, it will be complicated++.
Oh dear.
> Fortunately we have two very competent Albanian speaking communities
> keen to help out.
Good :-)
> > *extremely* good at detecting when A/V desync has happened; the only way
> > to reliably get things back into sync well enough is by using a
> > visualisation of the audio and to line that up with a visual cue that
> > can be easily related to part of the audio visualisation (like would be
> > the case with a clapperboard and the spike that its clap would produce).
>
> I've done this before with Blender so I understand what you are getting
> at.
Great.
> Somebody will need to identify the start points in the big WAV files
> and potentially cut them into smaller files and normalize them with
> sox.
Actually, if you're going to normalize files, it's better to use
bs1770gain. This normalizes the audio according to the international
BS.1770 standard of the ITU, which TV stations follow as well. If you
follow that, and you play the audio on a TV set, you shouldn't have to
adjust your volume setting (which you will have to do if you use sox,
especially since it "normalizes" on the highest spike and therefore may
be different for each video, or may be totally useless if someone at
some point during the talk coughed very loudly close to the microphone).
(also, it's best to do the normalization after the files were cut in
pieces, something that SReview will do for free)
[...]
> > Most nonlinear editors will support doing so. One nonlinear editor that
> > exists in Debian is "kdenlive". Not sure whether it supports that --
> > haven't tried yet.
>
> Blender is also in Debian and it is useful for this, although some
> people feel kdenlive is easier. I've read elsewhere that the Blender
> keystrokes are "wrong" for people who do this type of work regularly.
AIUI, blender is mostly a 3D animation package, rather than a video
editing tool -- but I wouldn't be surprised that it can do more things,
it's a big project.
> https://packages.qa.debian.org/b/blender.html
>
> https://danielpocock.com/quick-start-blender-video-editing
>
> > Alternatively, sreview (https://sreview.debian.net) is capable of
> > processing A/V desync correction values, but it doesn't have any way to
> > help you line up the audio and video yet.
> >
> > If you want to do the former, you'll probably have to do more work
> > yourself, and afterwards one of us will have to do the upload into the
> > meetings archive.
> >
> > If you want to do the latter, we'll have to get the raw files uploaded
> > to vittoria.debian.org, I'll have to add code to sreview so that audio
> > from alternate sources is supported (shouldn't be too hard, can probably
> > get that done next weekend or maybe a little earlier), you'll have to
> > give me the schedule, and then you and any volunteers you want to share
> > the load with can do review through the sreview webinterface.
> >
> > Let me know what you decide ;)
> >
> The schedule is on the wiki:
>
> https://wiki.debian.org/DebianEvents/ks/2017/MiniDebConfPrishtina#Schedule
>
> but from what you describe, it doesn't seem like sreview is the right
> solution for this type of raw data.
Well, I want it to be :-)
It's meant as a video review system for conferences. It's true that I'm
not quite there yet, but every new feature that I add will improve its
quality...
> However, I would really like to upload the raw data somewhere for backup
> purposes at the very least, can I still put it on vittoria.debian.org
Sure. That's what it's for, at least until the videos are published and
you're happy with the result.
> and if so, under which directory?
Your home directory would work, but you'll probably need access to
vittoria first.
> One other issue: I notice the DSLR doesn't always keep the speaker in
> focus when autofocus is enabled. It is a Nikon D800. Can anybody with
> more experience comment on this? Is there anything I can do to get the
> autofocus working well?
"flawless autofocus" is a utopia. I have a friend who is a professional
cameraman; he tells me pros never use it, they use manual focus instead.
It's the only way to reliably keep things properly in focus.
> Do I need to have the camera closer to the speaker for good
> autofocus? Or should I just put it in manual focus mode in future?
Probably :-)
> Are any other DSLR models better for this?
DSLRs are not optimized for video, and therefore their autofocus may not
be ideal for recording video. Nikon has 2 types of autofocus: the first
is the mode with external autofocus engine which was created in the
early 80s and is very slow; the second is the "SWM" mode, where the
autofocus engine is integrated into the lens (you need "G" lenses for
that) which is faster and less loud.
It's still slower and makes more noise than most video cameras, though.
If you stick with DSLRs, you could try different lenses, but beyond that...
We (the DebConf video team) have a long-term goal of being able to loan
out our video hardware to miniconfs that ask for it. We're not quite set
up for that yet, but next time we might be; do talk to us before going
to a conf.
Regards,
--
Could you people please use IRC like normal people?!?
-- Amaya Rodrigo Sastre, trying to quiet down the buzz in the DebConf 2008
Hacklab
Reply to: