Wednesday, March 02, 2011

PyCon 2011: Interview with Carl Karsten

Another Chicagoan making the drive to Atlanta is the man behind the PyCon videos, Carl Karsten of Next Day Video. After a discovery in 2008 at a Debian conference, he found a more productive video process that he took to every Chicago-area user group that would let him try it out, which got him to where he is with today’s PyCon video team.

The Chicago Python Users Group is one of those groups that Carl gets his experience with every month, along with local Java, Hadoop, Erlang, and Android groups. While local meetings like these are dwarfed by the three day conference that is PyCon, it’s a good proving ground. After a half-hour setup, all of the talks, then a half-hour teardown, it’s an encoding and checking party after that.

He’ll spend 30 minutes to encode one video to one format, multiplied by however many is necessary. “I am currently encoding to flv (because as flash is still king of Internet video), ogv (because html5 is the future king), m4v for iPhones and maybe other mobile devices, and mp3 for those that like to learn Python while they work out,” says Carl. On the difficulty, he quips, “it's quick ‘n easy, except when it isn't.”

As for a conference like PyCon, he’s looking at getting around 1.2 TB of footage to bring home. His record is to have a video online three hours after the talk was given, but it doesn’t look like that’ll be the case with 5 tracks and three days worth of footage.

While I knew Carl was a big proponent of open source, it was great news to me that he does 100% of his work with open technologies. DVSwitch from Debian developer Ben Hutchings handles much of the recording and produces the hundreds of raw files they take to post-production. That work is then passed onto Carl’s Veyepar library, which handles the various video metadata and uploading capabilities.

Use of a frame grabber allows Carl and the team to get a stream of the speaker’s screen, not just their presentation slides. Any code, examples, pictures, or video gets taken and mixed with a camera view of the speaker on stage. Since the grabber works off of the video feed, there’s no need for the speaker to submit talk slides in any specific format -- it’s all just video coming out of their own computer.

The whole post-production system is wrapped up with a Django app. From there he makes any of the necessary corrections like missing titles, incorrect scheduling, or any of the hundreds of little things that can go wrong in the process. “Most are recoverable, it just takes time,” he says of any hiccups.

For samples of what to expect for the PyCon 2011 videos, take a look at what the crew came up with in 2009 and 2010 at pycon.blip.tv, or take a look at any of his user group talks on carlfk.blip.tv. If you can’t make PyCon, hopefully the videos keep you in the loop!

1 comment:

Chris Church said...

Hats off to Carl for his dedication. It's great to see how this process has grown since I'd been briefly involved in trying to capture video a few years back in Chicago.