Geometry in Modal Homotopy Type Theory

Felix Cherubini

<2018-11-13 Tue>

Main workshop page

Recording and live-streaming of the workshop

The live streaming problem

To record talks, it would have been enough to get a camcorder and record to its SD-card or something like that. Live streaming (with zoom in this case, but the specific software shouldn't really matter) seems to be harder and it took me some time to figure out how it can be done. Figuring this out started with some disappointment that most camcorders can't be used as webcams via their USB cable. Also, "cheap" camcorders (<$3k) don't seem to be able to produce usable HDMI (or similar) output, so there is not even any way to get the video feed from the camera. At least if you want something like HD-quality, which I did since that is necessary to make the blackboard readable.

My solution (Specific to the resources I had - it might be easier to get some HD-webcam)

Fortunately, it is possible to loan more advanced camcorders at the CMU library. I chose to work with a Sony NX100. It produces HDMI-output that can be viewed on a TV. I did not have a computer with an HDMI-In (almost all Notebook-like devices only have HDMI-Out) and it probably wouldn't have helped anyway, since the HDMI would have been uncompressed. The cheapest solution for turning the HDMI into something like a webcam seems to be to get a video capture card. I bought something advertised as a "video game grabber" or something like that for $100 (specific product on amazon). Comments indicated that it is actually not good for recording HDMI-out of a PS4, but pretty good for HDMI-out with a lot less movement, like people giving talks with a blackboard. It came with the necessary cables and worked ok. It needs USB 3.0 - if connected to something slower, it will reduce the resolution or not work at all. It also worked only on one of my USB 3.0 ports and during the week I reduced from 1080p to 720p after it stopped working which (strangely..) made it work again.
Also, it lags a bit, so I decided to work only with the audio from the camcorder (which is included in the HDMI-out and processed by the USB capture device). This turned out to be somewhat of a problem, because the audio of the camcorder worked less well than during the test I made one week before the workshop. I would look into getting a lapel microphone which can give shoe/XLR input to the camera when doing this again.
Also, it might be a good idea to have a notebook with more CPU-power. Zoom seemed to use all it could get out of my i5-3320M most of the times I looked into it. The recordings didn't take much space before Zoom converted it to mp4 (<2GB per talk) and less after (<1GB talk).


CMU has some kind of corporate license that easily covered what we needed for the workshop. One "problem" I had was that I missed comments of Zoom guests once because the chat icon didn't show the unread message, since there was an open chat window somewhere that I didn't see.
More frightening was, that sometimes it took my computer/Zoom quite some time to recognize the camcorder/capture card. I don't know if desperately replugging everything helped.
One time, Zoom somehow decided (I don't know how I could have triggered it) to use the audio of the internal microphone of my notebook instead of the audio from the camcorder. Luckily it was pointed out quite fast in the Zoom chat, that something is wrong.


I needed my american phone to verify my identity to make google allow uploads of videos longer than 15 minutes.