-
Posts
23,180 -
Joined
-
Last visited
Content Type
Blogs
Forums
Store
Gallery
Videos
Blog Comments posted by Josh
-
-
9 minutes ago, aiaf said:
the size of texture can be larger on certain axis depending on the geometry on screen ?
There's a limited range, forming a square around the camera. Then you have another stage that is a bigger square around the camera, and then maybe a third. Right now I am testing a single volume.
-
1 hour ago, aiaf said:
Hmm when i look at the first video i think all that corridor should have been lighted up from the beginning.
Or at least more chunks loaded from the beginning in the corridor , not 1 by 1.
Maybe the stages should have different sizes depending on layout of level.
PS I have no idea how you render the lights and how expensive it is
There is a 3D texture used to store voxel data in, that moves around with the camera. That is what the video shows.
-
I've now got it working with multiple cascaded stages, but there are some big discrepancies between the lighting in different stages:
Even if some cross-fading is added to smooth that transition, that is still going to stick out quite badly.
- 2
- 1
-
It will be a relief, too. The audacity of what I have attempted to do is pretty extreme, but it all did work out nicely.
- 1
-
Yeah, I am kind of annoyed right now but I also realize this is the end of this long R&D era I've been in. After this last step, all the hard problems are solved. There won't be any new groundbreaking technologies to experiment with for a number of years, and all the focus will be on making good software. This is also nice, but it's just a big change in what I have been doing for four years.
- 2
-
-
I spent a lot of time chasing the sparse voxel octree dragon, but my final verdict on this is that Crytek is actually a generation behind what we are doing now. I played Kingdom Come and I don't believe their tech supports indirect specular, which is the most impressive aspect of this technique. They're probably hamstringed by console support, as we saw back when Leadwerks was the first game engine using deferred lighting.
-
-
Three at most. I want the default to be two. It was just hard to get multiple bounces working, now I can do as many as I want.
I think there will be a setting for framerate vs. latency, as I think that will be a place where a tradeoff can be achieved.
-
-
54 minutes ago, Genebris said:
Do I understand that correctly? Reflected visuals are geometry that got voxelized and then polygonized back?
The first part is correct, the geometry is rasterized / voxelized into a 3D texture. The reflection is a bunch of texture lookups into that texture.
-
-
-
33 minutes ago, Yue said:
GPU : GTX 1050 2 Gigas.
That's going to give very slow performance no matter what.
-
I noticed your performance was quite slow in your demo, but I'm sure it can be improved.
I would look at post-processing effects, and check to see how much the minimap is slowing your framerate down.
Also, how many particles is that smoke emitter using?
-
It's every frame now. It's very fast, but it uses a lot more memory than my previous approach.
- 1
-
-
Interesting that some of the screenshots in this article look very similar to the above:
https://www.gamedeveloper.com/programming/graphics-deep-dive-cascaded-voxel-cone-tracing-in-i-the-tomorrow-children-i- -
1 hour ago, IceBurger said:
With reflections and all these new rendering capabilities, I have to ask: would it be possible (feasible I should say) to make a house of mirrors horror video game with this?
It would not be that hard to modify the shader to reflect the ray and do a second bounce. Whether this is a good idea or not, I do not know.
- 1
-
4 minutes ago, IceBurger said:
Very cool stuff! Do you think UltraEngine will have Dynamic Resolution Scaling?
I'm not a big fan of this because it masks problems, but it is possible to do I think.
- 1
-
What I am finding is that anything above the second or third LOD just shouldn't be used for specular reflections. The cone step tracing technique should not be used for specular reflection, either. A more accurate octree traversal is the right way to go for this, and then the cone step should be used for diffuse GI.
-
-
There must be a formula to calculate this exactly. The non-linear depth value makes it tricky to figure out. I wish GPUs supported a linear depth buffer.
- 1
-
Yes, Lua will continue to be used.
- 1
Upgraded car physics (demo)
in Marcousik's Creations Blog
A blog by Marcousik in General
Posted
The drifting and hill climbing is really nice!