[Video] Current state of the art for conference recording

[Paul Wayper]

----- Original Message -----
From: "Tim Ansell" 
Sent:Mon, 17 Jun 2013 17:50:59 +1000

So I thought I'd quickly document what the current state is; 

	* TwinPac for VGA capture to DV,

	* Firewire based camera for presenter using DV,

	* Going into DVSwitch (pretty much as described
by http://dvswitch.alioth.debian.org/wiki/large_meeting/ [1])

	* veypar is used for conferences to allow quick processing of the
large amount of video, splitting into talks and uploading to
YouTube/pyvideo/other.

 	* Going into Flumotion or Icecast for Live Streaming (least tested)

 The current setup has been well tested and I believe well
understood.More or less.  My experience was that there were a few
people who knew about this system, but it wasn't well documented
online.  The documentation, such as the DVswitch page linked above,
is terse, non-adaptive and mainly handles the working case.
 Unfortunately, I wasn't able to speak with the people who really
know this system - Ryan and Carl - and some of the other people who do
know it - Tristan, James Iseppi, Matt Franklin, Euan, Liam and others
- were only able to help educate me to a limited extent.
The thing I think is really important in any system we come up with is
that it's well documented.  I know that's painful, I know that's
boring, and I know it takes a lot of time.  But it's essential to
being able to help other conferences and groups.
 However, the current setup has some limitations;

	* Currently only SD resolution

	* Getting features into DVSwitch is painful

	* TwinPac is End-Of-Life and impossible to purchase new anymore

	* Laptop's no longer have Firewire ports so ExpressCards must be used
and have a tendency to pop-out

	* Firewire Cameras are becoming impossible to find

	* Poor ability to integrate with other online systems such as
HangOuts, etc

	* Requires significant expert knowledge to run smoothly

 	* Requires 2 people per room.

 Have I missed anything? What do you guys think?
I second Euan's observation about Audio, but it does tie in more or
less with the setup as you've described.  The main limitations I
think can be added there are:
* Very difficult to get accurate information about what is actually
being recorded.  You can hear this in the variable audio levels in
this year's LCA videos.  This is also partly operator training, but a
system that gives you no idea of how the recording sounds or looks
cannot be fixed no matter how good the operator is.
* Standard configuration only handles one Twinpact and one camera.
 It's difficult to change that on the fly.
* Only records the final mixdown.  Premixed recording is possible but
it does not get the cut information.
On the other hand, I think two people per room is essential  One
person to operate the camera, and the other to monitor the feeds, keep
the audio levels reasonable, switch between video inputs, and mark cut
points in the recorded video.  I agree with Carl that it also means
that if one person absolutely has to leave the other can get by in a
pinch.
There are many things I'd improve in Veyepar, but OTOH IMO that should
be discussed separately.
There are also many things that caused problems for us at LCA 2013
that were nothing to do with the above setup, so I'm not mentioning
them either :-)
Hope this helps,
Paul
  

Links:
------
[1] http://dvswitch.alioth.debian.org/wiki/large_meeting/

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.linux.org.au/pipermail/video/attachments/20130618/3666f3ac/attachment.htm