Fool me, this is real good shit ! [the VR thread]

[Dvorak]

Yeah live sports will be a killer app.

Live p0rn.
Live *breaking news*
Live whatever.

Will totally outshine gaming if you ask me....

[pogo]

Yeah live sports will be a killer app.

Agree. Surprised they aren’t doing this already, putting a few 360° cameras court side or pitch side, on a drone or wire above, and letting you switch between them and look at what you want. Facebook have the money and server/bandwidth resources. Rights for top games are maybe tied up but pick something else, even an amateur event.

[gumboot]

Cameras in the ball!


[edit] OK yes that is meant to be a terrible idea, but despite that I do mean for stabilisation to be applied before broadcast, so you don't actually spin with the ball.

[edit2] Makes me think a motorised hamster ball of cameras would be a great toy/surveillance doohickey.

[edit3] In fact, remote controlled hamsterball live video streaming control would be a direct mapping to walk-in hamsterball VR controllers.

[gumboot]

Anybody with the time to do it [this is how I shirk] and a handful of cameras could prototype this, right? A trivial mouse-controlled web-app to stream video to skybox textures and to let the user spin that around with a mouse; and a server to capture several cameras worth of imagery, apply some static geometric correction, and stream it out. Also, something to look at from a distance (a distance so we can ignore intra-ocular issues).

Then someone with a DK can take the same data and stick their head right up inside of it.


I think Microsoft were demoing real-time fusion of multiple nearby viewpoints into a virtual camera about a decade ago. As I recall it was so you could have cameras left and right of a screen and synthesise an image of the viewer looking directly at a virtual camera in between the two (in the middle of the screen where you don't want a camera). If that worked (NFI) then I might assume that by now it should be possible for something like multiple points of view across the surface of a hamster ball to be fused into the two virtual cameras needed for two eyes. That might be a lot of compute, though, and you'd have to be doing your geometric correction way, way better than needed by the prototype above.

[pogo]

A remote control [hamster] ball for might be OK for a toy but not for broadcast streaming. But being able to control where you look would be enough of a killer feature for many sports. After all it’s what the cameraman does for you. These would be not only looking where you want but cheaper, being non-moving and not needing an operator, so there could be many more of them.

It would work better for some sports than others. Tennis would work well, with a small court and action taking place all over it. Football not so much, with cameras far away and high up to get a good overview, some with zoom to capture the action and for replays.

[Dio]

I've tried a bit of 3D film on the Rift and it's mixed bag. If you just stream straight to the VR, then the results aren't great because there's no motion. You can add some limited rotational motion compensation but A. it's not quite right and B. you just can't do positional, so it tends to provoke a bit of motion sickness.

It does give a very "you are there" feel but for it to be significantly more interesting than 2D the camera needs to be rather close to the subject; research indicates that beyond 20m pixels overlap anyway, so the amount of '3D' you get from it is very minimal. The low framerates and resolutions of current video are also problematic.

I'm not sure it will add a lot to, say, football, although watching goal replays from a close-up camera will probably be interesting - it could just end up being novelty value only.

One way to solve the problem is to do a virtual cinema and project movies onto a virtual screen, which then fixes the motion sickness and a lot of the framerate issues. This is apparently a really rather good solution for watching film, including 3D film - basically it's a 'virtual imax' and Blu-ray source material tends to be good enough (4K would be significantly better, especially but not only on a higher resolution headset). But then you lose much more of the presence.

[gumboot]

To be clear, the VR broadcast concept I have in mind wouldn't even attempt stereoscopy, and consequently wouldn't presume to cleanly handle anything nearer than infinity (which we now know to be 20m away) unless it was rendered locally. Eg., if you did have a 360° camera in a car, you might edit out the vehicle interior and render it back in with stereoscopically-correct detail.

Telehamsterballing is a bit bluesky, to my mind, simply because I don't know if there are adequate methods for implementing the virtual camera positioning needed to composite two virtual 360° eyeballs out of a sphere of cameras. I'm more comfortable putting the cameras nearer the middle and approximating monoscopy, but you will get within 20m of objects if it's a mobile toy you're controlling.

[gumboot]

Quick question for the stereoscopically-able who actually have access to hardware: What's it like with inter-pupil distance set to 0? Flat, I suppose, but does it break the whole effect when you still have head tracking?

And now without head tracking?

[Darkflame]

VR tours akin to this;
http://www.airpano.com/360Degree-VideoT ... ik_volcano
Would work wonders.

Other then that, a Google Street View app is, duh.

[gravy]

Good interview here, you get a pretty clear map of what OR are aiming for and what they want to do.

Didn't know the crescent bay was 90hz



Big stuff in there:

- Aiming to nail the sitting down experience first
- Feel this is the last headmount devkti they'll release
- But there's a lot of new things to try with user input

And they reiterate their commitment to games. For now

[Mqark]

TxK demo freely available - works well on my DKII. Looks lovely!

https://forums.oculus.com/viewtopic.php?f=42&t=19321

Thanks Jeff and Giles

[Dvorak]

Quick question for the stereoscopically-able who actually have access to hardware: What's it like with inter-pupil distance set to 0? Flat, I suppose, but does it break the whole effect when you still have head tracking?

And now without head tracking?

I just tried it.
Lowest valid IPD that could be set with the Rift Configuration Utility is 55mm.

Maybe it's possible to reach lower values through the SDK.

[Dvorak]

TxK demo freely available - works well on my DKII. Looks lovely!

https://forums.oculus.com/viewtopic.php?f=42&t=19321

Thanks Jeff and Giles

Where did all my smart bombs go?
This version is a bit harder than the XMas build, isn't it?

[rowan]

You can understand anything without experiencing it.

I'd add that experiencing something can change your understanding though.

[rowan]

You can understand anything without experiencing it.

I'd add that experiencing something can change your understanding though.