first end to end proof-of-concept of a vertex deformation animation system i'm working on.
the shader has 16 slots for deformation operators. it uses dynamic branching to dispatch each operator.
the beauty is that the operators can be dynamically reordered and duplicated. #unity3d
the shader has 16 slots for deformation operators. it uses dynamic branching to dispatch each operator.
the beauty is that the operators can be dynamically reordered and duplicated. #unity3d
January 3, 2026 at 2:10 AM
first end to end proof-of-concept of a vertex deformation animation system i'm working on.
the shader has 16 slots for deformation operators. it uses dynamic branching to dispatch each operator.
the beauty is that the operators can be dynamically reordered and duplicated. #unity3d
the shader has 16 slots for deformation operators. it uses dynamic branching to dispatch each operator.
the beauty is that the operators can be dynamically reordered and duplicated. #unity3d
fucking finally
the "undo" steps are actually independent operators. i'm not just, like, moving t back to 0.
i can put this on a torus mesh, unwrap into to cartesian space, apply deformations, then re-wrap to toroidal space.
fallback shaders see a torus. full shaders see a ~fancy~ torus
the "undo" steps are actually independent operators. i'm not just, like, moving t back to 0.
i can put this on a torus mesh, unwrap into to cartesian space, apply deformations, then re-wrap to toroidal space.
fallback shaders see a torus. full shaders see a ~fancy~ torus
December 18, 2025 at 1:25 AM
fucking finally
the "undo" steps are actually independent operators. i'm not just, like, moving t back to 0.
i can put this on a torus mesh, unwrap into to cartesian space, apply deformations, then re-wrap to toroidal space.
fallback shaders see a torus. full shaders see a ~fancy~ torus
the "undo" steps are actually independent operators. i'm not just, like, moving t back to 0.
i can put this on a torus mesh, unwrap into to cartesian space, apply deformations, then re-wrap to toroidal space.
fallback shaders see a torus. full shaders see a ~fancy~ torus
vertex deformation shaders are a vicious linear algebra skill check
December 17, 2025 at 1:11 AM
vertex deformation shaders are a vicious linear algebra skill check
how much rust do you have to write before you’re compelled to describe every project as blazing fast 🔥🚀
November 21, 2025 at 6:10 AM
how much rust do you have to write before you’re compelled to describe every project as blazing fast 🔥🚀
blender 4.5 nodes to create a truss. it samples a pair of curves and connects corresponding sample points with a cylinder. #blender3d
November 21, 2025 at 3:59 AM
blender 4.5 nodes to create a truss. it samples a pair of curves and connects corresponding sample points with a cylinder. #blender3d
i think ray marching on a deformed surface is completely ironed out now!
specular is pointing the right way, it doesn't fall apart with extreme deformations, and i can chain together several deformations. #unity3d #simulation
specular is pointing the right way, it doesn't fall apart with extreme deformations, and i can chain together several deformations. #unity3d #simulation
November 3, 2025 at 8:33 PM
i think ray marching on a deformed surface is completely ironed out now!
specular is pointing the right way, it doesn't fall apart with extreme deformations, and i can chain together several deformations. #unity3d #simulation
specular is pointing the right way, it doesn't fall apart with extreme deformations, and i can chain together several deformations. #unity3d #simulation
generalized norm mapping c:
November 2, 2025 at 2:41 AM
generalized norm mapping c:
i finally got this ~working: ray marching on a deformed surface.
ray marching requires a euclidean distance metric to converge quickly. the idea here is to deform the surface (ro) and the incident vector (rd) using the deformation's jacobian. the march then runs in a locally linear space. #unity3d
ray marching requires a euclidean distance metric to converge quickly. the idea here is to deform the surface (ro) and the incident vector (rd) using the deformation's jacobian. the march then runs in a locally linear space. #unity3d
November 1, 2025 at 9:35 PM
i finally got this ~working: ray marching on a deformed surface.
ray marching requires a euclidean distance metric to converge quickly. the idea here is to deform the surface (ro) and the incident vector (rd) using the deformation's jacobian. the march then runs in a locally linear space. #unity3d
ray marching requires a euclidean distance metric to converge quickly. the idea here is to deform the surface (ro) and the incident vector (rd) using the deformation's jacobian. the march then runs in a locally linear space. #unity3d
using nvidia's slang language + compiler to automatically differentiate a vertex deformation shader.
this lets me completely skip the process of manually computing the jacobian of a vertex deformation.
analytic normals without any of the legwork!
#Unity3D
this lets me completely skip the process of manually computing the jacobian of a vertex deformation.
analytic normals without any of the legwork!
#Unity3D
October 12, 2025 at 12:22 AM
using nvidia's slang language + compiler to automatically differentiate a vertex deformation shader.
this lets me completely skip the process of manually computing the jacobian of a vertex deformation.
analytic normals without any of the legwork!
#Unity3D
this lets me completely skip the process of manually computing the jacobian of a vertex deformation.
analytic normals without any of the legwork!
#Unity3D
I ported my water spectrum code to C# and made an automated build script. Spectrum generation, FFT pipeline setup, and hooking into the surface shader is all automated now.
The number of cascades, their scale, and all physical parameters are now configured in one panel.
#unity3d #simulation
The number of cascades, their scale, and all physical parameters are now configured in one panel.
#unity3d #simulation
September 29, 2025 at 3:04 AM
I ported my water spectrum code to C# and made an automated build script. Spectrum generation, FFT pipeline setup, and hooking into the surface shader is all automated now.
The number of cascades, their scale, and all physical parameters are now configured in one panel.
#unity3d #simulation
The number of cascades, their scale, and all physical parameters are now configured in one panel.
#unity3d #simulation
I added physically accurate viscosity term to my water sim. Cold ocean water typically has a viscosity around ~0.077-0.079 N/m. This video shows 0 N/m, 1 N/m, then 50 N/m. #Unity3d #simulation
September 23, 2025 at 2:29 AM
I added physically accurate viscosity term to my water sim. Cold ocean water typically has a viscosity around ~0.077-0.079 N/m. This video shows 0 N/m, 1 N/m, then 50 N/m. #Unity3d #simulation
NSFW/WIP. I'm really pleased with how the topology turned out on this asset I'm working on. I drew the seams with the grease pencil, then made a topology around them.
Making all the structures for a base to deform well, while also adhering to the seam design, is a fun challenge.
Making all the structures for a base to deform well, while also adhering to the seam design, is a fun challenge.
September 15, 2025 at 6:53 PM
NSFW/WIP. I'm really pleased with how the topology turned out on this asset I'm working on. I drew the seams with the grease pencil, then made a topology around them.
Making all the structures for a base to deform well, while also adhering to the seam design, is a fun challenge.
Making all the structures for a base to deform well, while also adhering to the seam design, is a fun challenge.
hand retopology study :) I'm getting a little better at the islanding workflow.
the goals were to have nice edge flow around seams, holding edges on all major features, and 100% quads.
two 6-poles snuck in, unfortunately.
the goals were to have nice edge flow around seams, holding edges on all major features, and 100% quads.
two 6-poles snuck in, unfortunately.
August 18, 2025 at 10:38 PM
hand retopology study :) I'm getting a little better at the islanding workflow.
the goals were to have nice edge flow around seams, holding edges on all major features, and 100% quads.
two 6-poles snuck in, unfortunately.
the goals were to have nice edge flow around seams, holding edges on all major features, and 100% quads.
two 6-poles snuck in, unfortunately.
Cloth! I added a cloth lobe using Estevez et. al. "Production Friendly Microfacet Sheen BRDF." I also wrote a little monte carlo integrator to bake energy compensation textures for image-based lighting. #Unity3d
August 9, 2025 at 1:14 AM
Cloth! I added a cloth lobe using Estevez et. al. "Production Friendly Microfacet Sheen BRDF." I also wrote a little monte carlo integrator to bake energy compensation textures for image-based lighting. #Unity3d