it’s very close to one yes. the structure on top is definitely from one of the levels, although i don’t know if the whole map is a replica of one of the scenes or just that part.
What advice/learning resources would you give to someone trying to start a project like yours, also how are you storing the voxels?0 (I saw in another vids comments that it wasn't an octree)
You can suppose that it takes around 104 bits (That's a lot. You still can squeeze it smaller if you change the int type into int_16 type.) if the voxel's body is built like this { int:4 bytes x; int:4 bytes y; int:4 bytes z; char:1 byte VoxelID; } VoxelId would point out to the voxel's data like it's sub-voxel model. Int means number and char means ASCII characters. 1 byte can contain around 256 combinations so it's pretty limited
The voxel grid is really only 128^3 tho, right? I mean you can see when you simulate the automata that what you claim are individual voxels collapse down to be the coarser, non-instanced voxel size Modern hardware can't simulate automata at the scale of 1 billion yet.. that's the equivalent amount of data of over 120 4k images. Either way, great work! I love the GI and art style you're going for!
Thanks, yeah that's what I'm referring to as sub-voxels. If you check my video "GPU Falling Sand", I actually have a compute based simulation, which can simulate up to 1 billion cells. Though I had to implement quite a few optimizations to make it that scalable.
you know shit is efficient when the stress test runs this well
Impressive. I like the vibe of super clean and sharp aesthetics
Now do 1 trillion.
Really like whatever shader setup you got going on
is that map from monument valley? looks kinda like it
it’s very close to one yes. the structure on top is definitely from one of the levels, although i don’t know if the whole map is a replica of one of the scenes or just that part.
all in the web browser :o is this with webgl or webgpu?
webgpu:)
@@xima1 i love webgpu
i want to eat it
There's maybe 50,000 there
What advice/learning resources would you give to someone trying to start a project like yours, also how are you storing the voxels?0 (I saw in another vids comments that it wasn't an octree)
I'm using somewhat an octree, but it's not a publicly available implementation
How many bits are used to store each voxel?
You can suppose that it takes around 104 bits (That's a lot. You still can squeeze it smaller if you change the int type into int_16 type.) if the voxel's body is built like this
{
int:4 bytes x;
int:4 bytes y;
int:4 bytes z;
char:1 byte VoxelID;
}
VoxelId would point out to the voxel's data like it's sub-voxel model. Int means number and char means ASCII characters.
1 byte can contain around 256 combinations so it's pretty limited
What software did you use?
This is a new engine that I'm currently working on
I'm feeling the stress.
It looks good 😊
oh wow...
The voxel grid is really only 128^3 tho, right? I mean you can see when you simulate the automata that what you claim are individual voxels collapse down to be the coarser, non-instanced voxel size
Modern hardware can't simulate automata at the scale of 1 billion yet.. that's the equivalent amount of data of over 120 4k images.
Either way, great work! I love the GI and art style you're going for!
Thanks, yeah that's what I'm referring to as sub-voxels. If you check my video "GPU Falling Sand", I actually have a compute based simulation, which can simulate up to 1 billion cells. Though I had to implement quite a few optimizations to make it that scalable.
hmm my 4090 running @ 300 Frames per Second in 4k. just saying
Isn’t that magicaVoxel