Slashdot Mirror
Creating 3D Environments Without Polygons
Igor Hardy writes "I've conducted an interesting interview concerning a new episodic indie adventure game series called Casebook. What's quite uncommon, especially for these kinds of independently developed and published productions, is that they include professionally created FMV — all of the footage is filmed in real locations. Yet what's even more interesting is that the games use an innovative photographic technology which recreates a fully explorable 3D environment through the use of millions of photos instead of building from polygons. The specifics of how it works are explained by Sam Clarkson, the creative director of the series."
It's an interesting approach but it reminds me too much of the old adventure games, I suppose it could work out all right if they make it fit in smoothly.
It'll be interesting to see what effect it has on performance though.
It's just photosynth but with an effort applied at hiding the individual photos all while turning it into a game?
Viable Slashdot alternatives: https://pipedot.org/ and http://soylentnews.org/
Something like this?
That's so eighties...
Do not click parent link, it is a shock site.
From the article:
So he's basically saving money by copying other people's work without licencing it?
Except that you get a smooth transition from one VR to the other.
A QuickTime VR - for those who have been living under a rock or just don't care - is a small file with a graphical representation of, typically, the whole environment. So 360 degrees around and 180 degrees up/down. Within a QuickTime VR viewer you can then look in any direction of that environment, zoom in/out, etc.
In some QuickTime VRs (and much better in older PanoTools-based panoramas, or even SmoothMove/etc.), you can click on a hotlink and it would take you to another QuickTime VR taken from that position/area (e.g. click on a door and you would get a VR of the next room).
This is much the same technology as far as that goes, except that instead of clicking (presumably), you move around using whatever you'd use to move around with typically.. such as the keyboard.
The nice part is where they blend smoothly between the panoramas. Sure, they have to take a LOT of them to begin with (hence the camera rig off a grid in the ceiling, probably something like 1 pano every 10 inches or whatever; from the looks of it only in a 2D plane, but 3D should be doable), but even with that you need some nice motion estimation to blend between the two panos as depicted on the screen.
However, there are limitations that they point out...
1. they can't blend in live actors -while- you move. That's an organisational limitation - you'd have to make the actor re-do their steps for every single pano vantage point. Ouch. You could mount a whole grid of cameras, but that's gonna be insanely expensive (not just in material costs but rigging that up for each room as well). Probably their best bet is to 3D digitize the actor and blend that into their panos using standard 3D compositing software.
2. they're limited to a 2D plane at the moment. As I mentioned, this could be made 3D - just means it will take a LOT more time to create
3. they're limited by storage media; granted, they're talking about their hope for a DVD release, so I guess they're stuck on CD, but even DVD or Blu-Ray would be filled up quickly if it was a more involved game than what it currently looks like.
This actually sounds like they are generating polygon-composed scenes from photographs. Cool, yes, but not actually without the traditional rendering method.
Of course, yes, it's possible to do this entirely with photos and without any kind of 3D rendering at all, but in that case, can it be accelerated? Will it move at a decent speed?
~ C.
A "photograph" is just a textured rectangle - i.e. a textured polygon. So the environment is created by the blending of many textured polygons. Sounds awfully familiar to me.
Sure, they are rectangles instead of triangles; and sure, they aren't arranged in a mesh. But this looks to me like the triumph of a marketing press release over engineering reality.
These posts are for a REPLY to my original post; it was trolled down and thus became invisible, making it look like my original post was the malicious one.
Srsly, do not click on the zoy.org link.
Okay, so these voxels - with current generation technology - are represented as cubes which of course are 12 tri-polies, so it's not entirely -without- polygons.. but at least it's not based on polygons and it lets you do some pretty cool stuff - such as truly fully destructible environments. No, none of that "we ran a script on all objects (except for those we don't want you to be able to destroy) that pre-fragments them and call the havok engine on the object if the damage model reaches a certain level" crap. I mean *fully* destructible.
Sadly, this is the only 'worthy' example I've seen and it's still kinda 'meh'... ...but it's got me excited for what could be done with voxel-friendly accelerators.
http://voxelstein3d.sourceforge.net/
The biggest hurdle, however, is not the accelerators... it's the artists. Suddenly you can no longer get away with 'modeling' a house by simply putting up a façade like they do in movies.. now that house has got to have an interior because some wise-ass IS gonna be spending all of the ammo on the map to chip away at that 10" thick brick wall so they can venture inside the house WITHOUT collecting the key from the bossfight.
So... shot my own excitement down there, again... but at least there's the potential :)
Quite impressive. Not much information how it works though.
http://casebookthegame.com/areo_rooms_demo.mov
www.casebookthegame.com
Right in the video they say they are doing Light Fields:
http://graphics.stanford.edu/papers/light/
So the lighting is captured by the camera, not an algorithm - how then, do you *remove* lighting for shadows? Or change the lighting when light-emitting objects move?
This seems like a step backwards from truly immersive worlds, where one can interact with the world and it interacts back. My prediction is that this line of research will lead to some cool proof-of-concept games (under a killing moon is still one of my favorite games of all time), but will ultimately be a dead end. We have the technology to do better than this for games.
Wouldn't it make more sense to base something on a volume particle system? You could start with only a few elemental particles ... say, three (you could get smaller but we're trying to get simple) ... and make up some rules about how they combine. make them up into, oh, say, 117 or so "elements" which you can then compound according to other rules. Each step in the chain can increase complexity.
Naw, it would never work.
09 F9 11 02 9D 74 E3 5B D8 41 56 C5 63 56 88 C0
People in above comments are talking about photogrammetry and voxels. This is not the technology refered to in this article. They specifically mention having to compress the photos to a great extent to get the game under 1 GB. I am 99% sure that what they are doing is simply storing a grid of 360 degree 'fisheye' photos, and then interpolating between them based on the camera position using some clever interpolation method. The technique is pretty obvious so I am guessing the technology they are so proud of is the interpolation method. This technique seems very restrictive to me, allowing no relighting or dynamic geometry. Its pretty much only good for this game. You can see in the video linked in a previous comment that there is no extra geometry in the scene when you are exploring it, not even your own legs ...
I couldn't plow through all the spam about the actors and characters and storyline, cut to the chase... is this just VRML-style backdrops like Myst and Riven again?
This is nothing new. That technique was a standard before the polygon technique became feasible.