Video Workflow

High tech: A coke-crate tripod on a ping-pong table.

Mike Zabrocki asked for some kind of run-down of how I’m building the videos for the algebraic structures notes. So here’s a quick run-down! There are currently two finished and five more in the pipe.

Video: An early Canon HD camera I picked up off Craigslist back in the spring; it’s actually really nice for a handheld. It’s mounted crazily to a big plastic crate pointing down at the table surface.

Video Editing: I then convert the video file to something manageable (interlaced mts files aren’t so useful to me). Then I use kdenlive to rotate, fix the color levels, and increase the speed by 300-500%. Render. The rendering step seems to take about 4 minutes per minute of video produced. Incidentally, kdenlive has been far and away the best video editor I’ve tried on Linux; good support for many formats, lots of useful effects and compositing built in, nice and versatile.

Sound: Then I watch the video a couple times to get timing more-or-less right, and record a voice-over using an H2 Zoom and a lavaliere microphone. The mic is great since it gets my voice and a very reduced version of the massive echo in the room I’m sitting in, the constant crowing of the chickens outside, and the occasional monkey wars that break out on the front lawn. The H2 is acting as an audio interface; as always seems to be the case, my laptop’s built-in sound stuff is noisy noisy noisy. Going through the H2 gets rid of all the line noise. (It’s also a truly fantastic little piece of hardware, generally. I bought another immediately after the first one blew out a couple years ago.)

Second Pass Video: Then I load the sound and video into kden again and render one more time. And finally…


It’s not exactly a fast workflow. I’m finding a reasonable way to go about it is to shoot a bunch of sections at once, then let the conversion script do its thing while I work on other stuff. The kdenlive video editing step is probably the slowest part, but then again, editing probably should be the slowest part…. The main thing is a bit of head-space for multi-tasking. Like writing the next few sections of text while waiting for the rendering to finish.