Well, the title is a bit of an over statement but it does seem that he is investigating the use of a 'sparse voxel octree' as a technology for the next ID software engine. Have a read about it here:
He even mentions carving up the world - so I wonder how dynamic it will be?
ConsistentCallsign at
Re: John Carmack is working on a voxel engine!
Voxels in Crysis and now voxels in an id engine. This trend is entirely consistent with this article's abstract: http://www.cs.sunysb.edu/~vislab/projects/volume/Papers/ said
The progress so far in volume graphics, in computer hardware, and memory systems, coupled with the desire to reveal the inner structures of volumetric objects, suggests that volume graphics will develop into a major trend in computer graphics.
voxels 1 - polygons 0
Edited by ConsistentCallsign at
Slang at
John Carmack said
(snip) It’s going to wind up having a depth value on there that you could intermix with anything else. (snip) You’ll probably still want to do characters as skinned and boned with traditional animation methods. While you could go ahead and work out a voxel method of characters using refraction skeletons around characters and you could do animation, you probably wouldn’t want to because we can make characters that look pretty damn good with the existing stuff and if everything continues to get 10x faster without us doing anything you’ll probably want to do characters conventionally.
That's exactly what I wanted to say! ;D
Edited by Slang at
bitshit at
He states raycasting through a voxel mesh is more efficient than "general purpose" raycasting, as you can take advantage of the spatial properties... Anyone knows what he means with that? You'd still have random memory access raycasting a voxel model?
IJs at
Slang said at
That's exactly what I wanted to say! ;D
Point taken. But I'd still be very happy to prove him wrong ; )
Hamtaro at
I think Carmack's idea is to have models still created as (fine) meshes and would still utilise the rendering pipeline in graphics card's gpus to go through vertices for his raytracing. I was imagining something closer to IJs' implementation. In stead of traversing through the voxel data in traditional sense.
(Hi, I've been lurking for a while, but it's my first post, so I would like to say this is a great forum to read about Ken's work and very interesting to hear about other people's implementations.)
bitshit at
Hamtaro said at
I think Carmack's idea is to have models still created as (fine) meshes and would still utilise the rendering pipeline in graphics card's gpus to go through vertices for his raytracing. I was imagining something closer to IJs' implementation. In stead of traversing through the voxel data in traditional sense.
(Hi, I've been lurking for a while, but it's my first post, so I would like to say this is a great forum to read about Ken's work and very interesting to hear about other people's implementations.)
Well he mentions he wants to raycast through the voxel octree, so it would require more flexibility than the traditional pipeline. He did mention he wants to leverage the traditional pipeline for things that are problematic with volume graphics; Character animation. Also these different pipelines could be combined in a hybrid renderer. Only difference between the method Carmack outlined and what Ijs is doing, is that Carmack only wants to raycast (first hit raytracing) to determine visibility, while Ijs is doing full fledged raytracing. Carmack argues the benefits of full raytracing... but currently it's only interesting words... While Ijs has actually something more interesting, a demo! :)
... Still have no clue about this ray coherency while raycasting volume data though...
ConsistentCallsign at
bitshit said at
He did mention he wants to leverage the traditional pipeline for things that are problematic with volume graphics; Character animation.
Animated voxel models can easily be deformed. It is problematic to interact with and shoot/inflict realistic holes/damage in/to an animated poly model.
Maren at
Everyone check Hamtaro's youtube now ;D
Spacerat at
Interesting is also the interview with Crytek's Cevat Yerli about Real-Time raycasting: http://www.pcper.com/article.php?aid=546
Zelex at
Carmack is going for a fully static voxel scene. No deformability or destructability.