At FOSSLC, we have been recording events for nearly two years now. We have hundreds of talks available online for free from our videos library. Along the way, as a non-profit organization, we've had to get ruthlessly efficient in how we do things. The technology used to record events is a good example of this. I would like to share the history and progression regarding how we drove down the cost and size of our video recording solution by an order of magnitude in both dimensions. We'd also like to announce a new open source project. There is an audience participation aspect to this post because we need your help to name it. Read on.

Prehistory - how we got started

We started out running events teaching skills with open source software. In the beginning we ran these events without recording events. The crowds averaged 80 people per event, but once the event was over, the content was lost. It occurred to us that if we could record the talks, we could reach more people. Thus the vision was born to record. We decided to experiment to see what happened.

Version 0.x - Early days - a vision to record great open source talks

We evaluated our options to record talks. The pickings were slim. While attending an event at Carleton University about ePresence I met a firm that did this. Thus we started paying a company to record our events. Pricing ranged to record the talks but was more than we could afford since the costs to record were approximately 70% of the costs to do an event. The solution this firm used consumed a 1m(3ft) by 2m(6ft) worth of space when running and took an hour to set up, a half dozen people to haul in and out of venues, and the encoding only ran on  Microsoft Windows. Thus it ran counter to our reason for existing both in terms of open source use and efficient use of funds.

Here is an example of what version 0.x produced.

Version 1.x - Cheaper, portable, running on Linux - the mobile capture station

We then worked with ePresence to develop a mobile capture station that would be smaller, run on Linux, and cost less. The results were promising and we were very impressed with how quickly the team at  ePresence developed the solution. The solution ran on a Linux laptop, fit in a backpack, and took roughly 10 minutes to set up. This was a dramatic improvement. The list price for the software was roughly on par with what it cost us to record 2 events before so we'd get unlimited events for the price of 2 events previously. Life got better.

Here is an example of what version 1.x produced.

We wanted to record OSGeo's FOSS4G conference and the  FSOSS conference. These conferences ran in Australia and North America at the same time in October. Thus, we needed more gear to pull this off. Despite generous support from our sponsors, the kind of funds needed to pay for a handful of new ePresence licenses were not available. We also learned that ePresence changed their license and were no longer open source as they were previously. This gave us significant heartburn as we like to support small businesses, and we valued our relationship with ePresence. We couldn't get the job done based on this thus we had an "itch to scratch".

Version 2.x - Faster, cheaper, open source, open standards

I challenged an talented intern whom had formerly worked for me at Ingres, Google Summer of Code, TalentFirst Network and who probably attended more FOSSLC events than anyone (hi Thanh!). The goal was to experiment to see if we could record video from the epiphan vga2usb device and audio from the sennheiser wireless microphone using strictly open source components from the command line. Without too much pain, it worked and Thanh quickly tuned it to work well. The next task was to create an easy to use gui. For this, we chose PyQT - the python bindings for the QT widget set. Using Python made sense for portability and very rapid prototyping. Using QT made sense given it's richness and liberal license.

As a result of this software, we can now equip 4 sites for the price of  a single site running with version 1.x. This is roughly 2 sites equipped for unlimited talks for the price of a single event based on version 0.x! What we find most exciting is that the cost of the gear for 2.x is roughly on par with the registration fee for most open source conferences. We've been working with open source conference organizers to let FOSSLC correspondents in to events for free if they are recording. We thank the conference organizers for supporting this and it seems to be working to capture even more great talks.

While we have clearly given up full motion video of the speaker for now, the video of the screen is bigger and full motion video by default. Thus debatable that this is in fact superior. You decide:

Here is an example of what version 2.x produced.

We used this new software for recording FSOSS and OSGeo's FOSS4G conferences. We recorded well over a hundred talks with our new software as of this blog post. While not perfect, we have investigated and solved the known crashes and quirky issues thus far. Our success rate is over 90% for capturing talks. It's a good time to invite others to join us to develop the software, or use it to record great open source talks.

In order to make further improvements and not require special software on most major Linux distributions, we switched to gstreamer for our video engine. It has done an excellent job.

The following is an example of the output from this version. We feel it is the best yet as of the time writing this post.

Thus a new open source software project is born. One that records video from any vga output. This is significant since often conferences recorded using a camcorder (as perhaps another cost effective solution) result in horrid sound, and crappy video. Personally, I find watching camcorder talks gives me a headache. With this solution, you get an excellent recording of what ever is being shown on the screen - even full motion demos, and the speaker's audio nice and crisp with good audio levels.It is almost as good as being there.

This software features:

• An easy to use gui with a start/stop recording button and text boxes for the speaker & talk title
• Real time thumbnail image of what is being recorded from the screen
• Real time audio level feedback (so you know the sound is ok)
• The speaker, title, and date recorded is automatically added to the file name
• The video file output can be transcoded to nearly any format using ffmpeg
• The software is free for use, free to distribute to as many sites as we like
• The software is open source and thus the costs, risks, and benefits of developing it are shared with any who chose to do so

We need your help

We will be making the code available from github (obviously with an open source license). We will announce the mailing list, wiki, and bug tracker shortly. Before we do... we need your help. What should we call it? Please comment here to make a suggestion. We'll create a poll to let people vote on what the development team feels are the top 5. If you provide your email for us to contact you, the winning name will get a very nice golf shirt with heat dissipating/ sweat wicking /bacteria killing technology (value around $40).

If you are interested, we encourage you to get involved with this project by using the software, providing ideas for useful features, reporting any bugs, and possibly contributing code should you desire.