Hello!
The thing bothering me most (and I know I'm not the only one) is about missing eye contact.
Cam is mounted on top of monitor or - best currently available - in the frame itself. You have to concentrate to look into the lens but even if I'm able to do it - and I fail a lot - it just feels weird.
Q: Is there a solution for this?
Don't tell me I'm the first one asking this! Idea: Some second/third cam streaming into PC and processed by an (almost) almighty GPU doing something related to (late and lamented) light-field cam.
Result: I look into the eye displayed in the monitor during the conversation and on the other side: vice versa (sorry, lame pun).
I did a zoom meeting for a class presentation where I used the 5D2 in liveview with the program ''camera live'' on my macbook pro.
I tried placing the 5D2 just above the screen so that its easy for me to look into the lens while talking. I just slightly adjust my gaze downwards to look at the people talking on zoom. I reduced the zoom window a bit and pushed it towards the upper portion of the screen as well to really reduce the elevation change of my gaze and it worked quite well !
Webcasts are a tricky thing honestly. Listening to people I dont think your gaze matters much. I think you gaze while talking is much more impactful for ''connecting'' per say.
There is a solution (https://en.wikipedia.org/wiki/Teleprompter), but...
(https://upload.wikimedia.org/wikipedia/commons/thumb/2/20/Teleprompter_schematic.svg/625px-Teleprompter_schematic.svg.png)
(https://upload.wikimedia.org/wikipedia/commons/thumb/9/90/Teleprompter.jpg/421px-Teleprompter.jpg)
Thanks, Ant123!
Found some interesting links there. Haven't heard of "Interrotron" before, for example.
Anyone having something heard/read about a solution involving an array of smaller cams attached to conventional computer monitors?
If you place 2 cameras on the left and right side of your monitor, sync them (not sure how), stereo-match their images and compute the middle view, could this work?
Optical flow could be used for morphing each of the two images into the other, with intermediate steps:
https://www.magiclantern.fm/forum/index.php?topic=20999.0
That particular procedure is not fast enough for real-time video, but it can probably be sorted out (lower resolution, GPU implementation etc). If you can capture (or find online) two images from left and right, ideally with similar FOV and DOF, I can compute the middle image to see if the approach has any chances of working.
I place a camera in the middle of my two screen set up.
I like using a longer lens and putting the camera further away, with a second monitor directly below the camera. I place the participant window right below the lens. The distance makes it hard to tell I'm not looking right at the center of the lens, considering that your eyes move around when reading a teleprompter but you can't tell unless the camera is too close to the subject.
This seems to be a good solution !