CS 473: Video setup

Last year I taught my first algorithms class in a room that did not have video-capture hardware built in, so I recorded the lectures myself. The resulting videos were arguably of higher quality than past semesters' videos captured with the built-in equipment. So even though I'm teaching in a room with video-capture hardware this semester, I'm likely to record the lectures myself, using the setup described below. Thanks to several people for input, especially Matt West and Luke Olson.

My setup is considerably less sophisticated than Erik Demaine's, and this shows in the quality of the resulting videos, but I'm deliberately aiming for simplicity and convenience over quality. My main design goal is to do everything myself with equipment I can easily carry in my laptop bag: a Macbook Pro, an iPad Pro, a power cable, a Lightning-to-USB cable, and an Apple Pencil. Altogether the in-class setup and later postprocessing adds only a few minutes of effort.

I use an iPad Pro as an input tablet, with an Apple Pencil, running the Notability app.
I tried lots of other note-taking apps. None of them do exactly what I want, but Notability is the closest approximation.
Notability works particularly well with the Apple Pencil (but not so well with other styluses).
I don't recommend this setup with any other iPad model. Despite the high-resolution Retina display, the input resolution on most iPads is poor. Even using a crappy stylus, the Pro is noticably better than previous models, and the Pencil only works on the iPad Pro. But I hear good things about Wacom tablets and the Surface Pro.
The physical setup would be much more comfortable with a raised and tilted lectern/laptop stand to rest the iPad on.
I run Quicktime Player to mirror the iPad display on my MacBook Pro, which is connected to the iPad by a Lightning-to-USB cable.
I would much rather mirror the iPad display using AirServer, but my universty's wifi network has disabled multicasting. On the other hand, the wired connection means I never have to worry about running out of battery. For similar reasons, always plug in the laptop.
Quicktime Player records the mirrored display to create the video.
Notability has a screen-recording feature, but it only records when objects in the final version of the slide are created, as if they were created in their final locations, scales, and orientations, along with audio. Objects that are drawn and later erased do not appear int he video; objects that are drawn and them moved are "drawn" in their final positions only. There are also battery and memory capacity issues with recording 75 minutes of video directly on the iPad.
The MacBook is connected to the classroom projector by an HDMI cable.
Or a VGA cable. Or wirelessly to the Apple TV. Or whatever. The display infrastructure is actually built into the room, so it works seamlessly.
I wear a wireless microphone to amplify my voice in the classroom. The audio is recorded by the built-in microphones on the laptop, which is placed as close as I can to one of the classroom speakers.
Yeah, this is an embarrassing hack, and the resulting audio sucks. (But it's stil better than the audio that Echo360 produces.) A better solution would be to use a wireless mike with an output splitter on the receiver, with one output going to the room speakers and the other going to the MacBook for recording. Working on it.
After class, I post the videos to the course website as downloadable MPEG-4 video (AAC, H.264, 480p)
Downsampling the raw videos to 480p via Quicktime Player takes a few minutes on my laptop; to make the resulting video streamable, I just change the .mov extention to .mp4. (Unfortunately, I often forget that step.) Uploading the 480p video to the course web site takes a bit longer, especially over wifi. Fortunately, both downsampling and uploading can be done in the background. On average, each 480p video takes up about 1.2Gb.
Finally, when everything is ready, I add direct links to the lecture web page.
I keep a local copy of the entire course web site (including the downsampled videos) on my laptop. I edit the raw HTML files using SubEthaEdit, and then synchronize with the course server via sftp using Transmit.
When I started recording lectures myself, I published videos on two video-streaming servers: Echo360, run by Engineering IT (like most other recorded CS classes), and Illinois Media Space, run by campus Technology Services. Despite some advantages of using these servers, the additional effort compared to just posting the raw files really isn't worth the trouble, in my opinion.
The biggest advantage of MediaSpace and Echo360 is that they have dedicated file servers designed to store and serve streaming video; the standard course web server does not. Downloadable videos eat a lot of space; I estimate that my videos this semester will consume about 15% of the disk space that was free in January 2017. (On the other hand, disk space is dirt cheap.)
On the other hand, uploading and labeling videos on MediaSpace and Echo360 takes significantly longer than just making them available directly. In particular, Echo360 needs several hours to convert one 70-minute video into its native video format, and the resulting audio sounds (to quote one student) like an over-amplified 1960s long-distance telephone call from a Dalek. MediaSpace is more efficient and friendlier, but it's only accessible with a university NetID, it doesn't allow downloading, and it collects analytics (who watched which videos for how many seconds) that many instructors may find useful but I find offensively creepy.