Video publishing supports more learning styles

Publishing course videos supports more learning styles and doesn't have to affect privacy.

Part of a series on the Future of Teaching

What if all the talking in a course didn't disappear right after the course was over?

When we went online, many people thought: avoid recording courses, that's a privacy risk for participants. I firmly think this is the right choice: I don't think any privacy risk to participants is worth it, and "don't say anything if you don't want to be recorded" isn't good enough, either - I don't want to push "publish" and have to hope that no one missed the warning. I don't want to motivate participants to be silent. Editing videos takes a long time and is hardly worth it.

This is part of why we developed livestream teaching: we want to separate the instructor interaction from learner interaction, so that there is no privacy risk whatsoever when recording. This only works if the livestream is engaging enough, but our previous posts show how we handled that problem.

In order for a video to be useful, it has to be published quickly. Watching videos months later isn't that engaging[1], but as a immediate follow-up for things you missed, or catching up if you had to miss a day, it is extremely useful. We can't have a long publishing process with this.

So, with livestreaming, what do we get?

• The livestreaming platform usually records the video, making it immediately available in raw form. This usually gets a lot of views, even if it is raw.
• Extensive editing isn't needed, since you aren't looking for privacy issues in the stream - just making it "good enough" in the amount of time you have.
• Learners can catch up immediately or refresh themselves on what they saw going off into the future.
• If learners know videos will be available, they are suddenly much more free to go with the flow of the course.

We actually made our own tool, ffmpeg-editlist, that allows us to define cut points in YAML file, and then run a process to do the editing. This allows us to distribute the editing via git, and copy-and-paste from previous years to save time. Thanks to this, it's our standard to have videos published by midnight the day of the course.

Overall, this works well. We seem to get lots of views with the Twitch automatic video (which lasts for 7 days): the same day as the course, usually 1-2 times the number attending the livestream (stats from Python for Scientific Computing 2022). The YouTube videos tend to get much fewer, since it's not ready on time for people catching up the same day. I'm still the main one making the videos, but it's simple enough that others could do so. I think I put in too much effort, and if I wanted it could be much faster - say, take only an hour per day.

I wouldn't recommend everyone try to make perfect videos for everything, but it's a nice advantage of livestreaming, and if you want text-based video editing for other events, ffmpeg-editlist might make it possible.

In short, I don't think the point of video publishing is to make a high-quality standalone production (although we can do that, and it can work well, especially with co-teaching). The most direct impact is supporting diverse teaching styles in the short term.

Read more:

• video demonstration of ffmpeg-editlist
• Video editor role description in the CodeRefinery manuals
• ffmpeg-editlist
• Sample ffmpeg-editlist file, very well done (perhaps too much).

[1] Before remote teaching in 2020, an argument against recording the teaching was "it won't be interesting for others to watch later". This post also shows how that's the wrong perspective: the videos aren't only for random people later, but people in the course already.