Sramana Mitra: Are you assuming that there are three parallel streams being picked up? Where is the mixing happening? If you say there’s no post-production, is the technology determining how to do the mixing or which stream to use in the edited version or are you showing all three screens in different windows?
Eric Burns: It’s a bit of both. One of the things that we tried to do is infer as much as we can about the content that should be shown based on the instructor’s behavior. For example, if you’re presenting PowerPoint and you begin capturing your screen, we’re going to automatically switch and show you the screen even if the slide is still up. Similarly, if you add a document camera halfway through your presentation, you’re probably going to show something on that. So, it’s safe to switch to that.
On the other hand, it is possible to capture all of these streams and a human editor, in a post-production, can say, “Show this stream at this time and this stream at this other time.” If you make sensible choices about how to tell the system what to show when, you can guess right most of the time.
One of the things that makes Panopto fairly unique but not completely unique is that we don’t compromise between the choices of streams. One of the things that goes wrong with the recordings of TED talks and a lot of online lectures is that the camera will occasionally switch away from something that you were looking at. Maybe you’re reading a slide and it cuts back to the presenter or you were really hanging on to the presenter’s words and this dry, boring slide comes up. We actually show the students both of these at the same time. We show the slide or the whiteboard side by side with the video. Just being able to show two at once removes a lot of these what-should-you-show-when questions. When you have multiple sources, either we guess right or the editor can do it in post [production].
One of the most interesting aspect about this technology is in situations where you have many streams of data all of which are relevant at the same time. For example in the operating theater of a teaching hospital, you want to capture half a dozen monitors, several different cameras, and multiple audio tracks, but there are not many systems that are capable of doing that. Panopto is probably the easiest to implement and cheapest. This is not just an abstract research project. It has real applications for high-end visual learning.
Sramana Mitra: In terms of the classroom infrastructure, I guess you have classrooms fitted with cameras at right angles and computers set up to capture things and you just teach the instructors to click record and it starts going. Is that how it’s set up?
Eric Burns: That’s pretty much it. There are a lot of different ways to solve this problem. One of the interesting things about going the route of building only software and using whatever commodity hardware the marketplace delivers is that you’re not casting your lot with a particular hardware form factor. We work with schools who do everything from remotely roll out an automated recorder to every podium in the school. This automatically records on Microsoft. We had a school that did this and they wrote a blog post for us called, Deploying Lecture Capture Overnight. They literally mean overnight because one day, it wasn’t there. The next day, there were capturing every class.
That’s one extreme end of the spectrum and it goes up through instructors using their own computers and webcams to, as you mentioned, fix installation cameras, videographers transporting cameras to classrooms using low-end webcam to the million-dollar room where you go for the high-end production. One of the drivers of success for the system has been that because it’s been generalized, it adapts very well to the unique capture cases that each school develops to match their particular flavor of learning.
Sramana Mitra: What does this cost?
Eric Burns: It’s licensed on a per student basis at scale for education, and on a per employee basis for enterprise. If you’re talking about a very small school with only a few thousand students, a small installation can be in the $10,000 ballpark. This can scale up to the $250,000 a year for very large institutions.
Sramana Mitra: Why is it priced by students and not by how many rooms are being equipped with this lecture capture technology?
Eric Burns: That is an excellent question. We don’t sell creation tools, a library, or a delivery platform for video. We sell one system that handles all of those functions where you can get from point of acquisition to point of delivery in Panopto. If you look at a system like that, where does its value come from? It comes from the amount of content that’s being created. If you want to discourage use of the product, the fastest way to do that is to assign a price to the number of recorders that you roll out and you are, in effect, putting a usage tax on the amount of content capture.
On the other hand if you price at the student level, you are encouraging open-ended content creation, which works very well if you have deployed servers where the school can buy their equipment and use their own network. If you model things that way where the student is paid for – now you have to recoup your investment in the form of capture – you wind up encouraging massive amounts of capture. For example one of our customers – University of Essex in UK – is capturing 80,000 hours per year of classroom content. They can do this without any regard for cost because they’ve paid at the student level.
Sramana Mitra: Good answer. Tell me again what year was it when you started this company with venture capital.
Eric Burns: The official date of founding was, I believe, May 2007. We didn’t really get going immediately. I didn’t leave Microsoft until September. At which point, I stepped in as the CTO and began recruiting the team.