Hacker News Reader

Comment by petra | Hacker News Reader

Mounting the camera behind the screen could create eye-contact in video chats.

That's the biggest difference between video chat and face-to-face conversations, and it creates low trust.

And while xiaomi isn't the first to achieve that[1], xiaomi is a consumer company, and their price point is very affordable.

So i hope they target this problem. they could be very successful, with good implications(environmental, social), hopefully.

[1]https://www.prnewswire.com/news-releases/dve-launches-camera...

View on HN

> Eye-contact in video chats. That's the biggest difference between video chat and face-to-face conversations

This issue is pretty far down the list IMHO. Above it are fixing the obligatory 5 minutes of "Can you hear me? I can't hear you" before the actual conversation starts.

Face-to-face conversations also lack "Please mute your mic when you're not talking because the background noise is annoying". "Did you forget you're muted?". "Can you repeat that, I can't understand you". "The video/sound is cutting out". "Where's this echo coming from?". "Please say something on your end". "What's that noise?". "Please use a headset next time". "Sorry, the call dropped out". "Let me call you again, maybe this will fix the issue". etc. etc.

Video chat is one of the great unsolved problems of computing. Yes, camera location is there as well - especially on Dell laptops that for some reason have it looking up from below into your chin. But that is really one of the lesser issues as far as my experience goes.

Sounds like you’re limiting your view to work-related calls. If I could have real eye contact on personal calls, it would make a huge difference. My current technique of looking just below the camera already makes a huge difference, and it’s only a hack to make it seem like eye contact for the other person.

You would have to look directly into the camera, no? Why just below?

I would guess because if you always looked directly into the camera, it would be harder to watch the video/see the other person at the same time.

It is natural to look at the person and eye contact when talking to them. Hence we don't look at the camera, though I'm sure some people can and do this.

I do half of that crap on personal calls too. Interestingly enough, I don't get those issues on phone calls! Or most audio only calls over the internet.

The underlying issue is the technical failures of video calling, not the social environment it's used in.

On laptop I would reduce the video window to 10% of screen and put it just under the camera and it made a huge difference

I've had decent luck tilting the laptop screen down a bit so that it reduces the distance between the camera and the window

> Above it are fixing the obligatory 5 minutes of "Can you hear me? I can't hear you" before the actual conversation starts.

Whenever I’m in this mode (which basically means whenever I’m on a conference call) I wonder what percentage of the GDP this represents, in aggregate.

Supposedly, way way less than the portion of the GDP that is made possible by remote work.

Those problems are solvable, the solutions are here.

And there are even more problems with consumer telepresence:

http://hellarddesign.blogspot.com/2013/07/why-living-room-vi...

But i feel that they too are solvable, if you care enough(like when you want to keep in touch with you senior parents).

Senior parents might not be as forthcoming as a project lead about the fact that the call quality sucks and that they can't hear half the things you say. Just a thought.

I spent the past 5+ years in jobs where I had to have multiple video calls per week. I used every tech imaginable, from those that cost five figures to the gratis ones. If these problems are solved, why is the experience universally terrible on all of them?

When I call my parents, it's over the plain old phone line, because I actually want to talk to them. Not seeing their faces on the screen is a small price to pay for not having each and every call turn into an impromptu debugging session.

What about a microphone array ?

Have you tried putting that both on your and your parents' locations ?

And what kind of internet you two are using ? and are the devices using wifi , or are wired ?

I have the opposite problem. Call quality over cellular networks is always terrible. Even when you can hear each other, the quality is awful. I generally use voice over internet with something like signal.

I also find that Apple's FaceTime is the most consistently good and functional with little "can you hear me now".

> "Can you hear me? I can't hear you"

Would you mind explain me why there isn't anything (or only little) done in that area? I guess we could implement some assistive checks for input and output audio streams? What is blocking us to implement a "handshake" that will verify that all is fine? - let's say communicators will send a recorded sample "Hello World" message over wire and check the microphone, volume, noise etc. In my opinion the pattern how we resolve the "I can't hear you" problem is a repeatable checklist that we more or less confidently do, so why we can't implement it as some algorithm that takes care of those problems in automatic way?

because your microphone can't hear what's happening in your headphones, and your computer doesn't know if you're using headphones.

A system that has access to both the microphone and the headphones could coordinate both based on past experiences (e.g., learning). I don't know how much data you would need to make it reliable though. If you know how the data is going to be transmitted, and assume little network delays, recording yourself before a video session might be of great help. Probably the same problem that TV broadcasters have. It's hard work to get quality real time content.

Sounds like you've only used shit systems on shit networks.

Filmmaker Errol Morris adapted a teleprompter to achieve eye contact with interview subjects. The screen is horizontal, below and in front of the camera, an inclined 2-way mirror moves the screen image in front of the camera.

http://www.whiterabbitdesigncompany.com/uploads/1/4/1/3/1413...

His non-patented "Interrotron" inspired the patented "Eyedirect". Other variations may be possible for small batch manufacturing.

Yes, eye-contact in video chats is very important. The way this was done in the early days [circa late 1990's and early 2k](probably still today in more elaborate ways) in video conferencing was the use of a one way mirror at an angle, camera behind mirror and below the mirror the display screen projecting up. This presented the display and the camera at the same level as the eye level of those upon the display, making for a much more comfortable engagement. By that, have a chat with somebody and look at one of their eyebrows, this makes them uncomfortable and increases the number of times they blink as well. The difference of eye-level contact in video conferencing/chats makes the experience and interaction around the subject and not limited by the technology.

EDIT - the same trick for eye-level camera positioning is used for some autocue setups in Television studios.

> it creates low trust

Try looking into the camera when you speak rather than the screen.

Then they'll trust you, but you won't trust them

+1 I do this when doing 1-to-1 or 1-to-many calls. I look at the lens at the top of my laptop or the one in the meeting room when I am talking.

It means I can't see what others are doing and feels kinda weird just staring at a lens when talking, but in a world of people who don't make eye contact like this on video calls, I think it has a good amount of impact/gravitas.

Wouldn't that require the camera to be located nearer to the middle of the screen, where the on-screen eyes of the other party would be? I wouldn't mind a slightly perceptible camera dot at the top of the screen, but if it was in the middle it might appear pretty conspicuous.

It will need to move. On anything bigger than a phone, the video chat may be in an off-center window. And i guess multi-window group chats are a problem too.

A great solution for a very limited problem.

well, that will require making a hole in the battery

What we talk in the industry as a new de facto standard is called "1B 1C layout" — a layout with 1 board and one battery cell edge to edge. A maximum by how much you can move the camera is determined by how much you are ready to sacrifice battery size.

It also has the eye-contact with one's self aspect, think of people who use their phones for doing their makeup as well as doing selfies.

A lot more of this goes on than video conferencing. Culturally the eye contact thing is not actually universal, there are understandings that the lack of eye contact is due to the video conferencing setup. So maybe the personal vanity selfie taking use case is what wins it here.

Does the screen have to be off where the camera is located?

I haven't dug deep but the few videos I saw showed a black bar covering the camera while it was in use (and bright colors when not).

I guess depending on how fast the camera is they could turn off the screen for a fraction of a second in between frames to take a picture. Not sure if that's feasible atm though.

That could actually work. They already make low persistence OLED displays for use in VR, where the screen stays black for most of the frame and a bright bar of image rolls down the screen.

The iPhone can already operate in this mode.

I assume there is a display brightness penalty, but the image quality from the camera should be fine as long as it gets more than half the frame collecting light.

I think it's quite real with OLED. But even with screen on, if you use a specialised CCD with huge dynamic range, I think it's doable.

You might find this interesting https://ieeexplore.ieee.org/document/1521517

A software fix seems more realistic. The latest smartphones packed enough AI circuitry to be able to infer how an image should look if you were staring into the camera.

> Lack of trust due to lack of eye contact

> Let's use fake eyes instead

Not sure that works out there..

How do I get to where I and someone else can trust each other and spend time in person? I am never going to want to perform in front of a camera. I refuse.

I don't think it would accurately recreate eye contact. Maybe some kind of uncanny valley like version of it, which seems worse than what we have.

It actually does, if you get a chance to use it (I have). Nothing uncanny about it whatsoever.

Keeping the person's eyes roughly around the camera area is achieved by moving/zooming the video on your screen to the appropriate spot. People don't usually move their heads that much, and when they do you can manually or automatically readjust as required.

Reply on news.ycombinator.com