A non-Lambertian surface is a material whose appearance changes with viewing angle, because it reflects or refracts light instead of scattering it evenly in all directions. Glass, mirrors, polished metal, and glossy plastics are common examples. The color seen at any point depends on the camera position, the lighting, and the surrounding scene.
Most 3D capture and rendering quietly assumes Lambertian reflectance: the idea that a surface scatters incoming light uniformly, so a point looks the same from every angle. Matte paint, paper, and unfinished wood behave this way.
Non-Lambertian surfaces break that assumption. A mirror sends light in one dominant direction (specular reflection). Glass bends light as it passes through (refraction). Brushed metal and fabric reflect differently along different axes (anisotropy). In each case, the observed color at a single surface point shifts as the camera moves, because what you are seeing is partly the environment reflected or refracted, not the surface itself.
This view-dependence is why these materials are difficult for systems that try to recover one fixed color and one fixed position per point.
Non-Lambertian materials are where high-fidelity 3D reconstruction tends to fail. Reconstruction methods detect and match visual features across photographs, but on reflective surfaces those features are virtual: a highlight that moves with the camera rather than a fixed point on the object. Triangulating these moving features places points in front of or behind the true surface, producing false geometry.
Recent research formalizes this. Dubar et al. (IEEE Open Journal of Signal Processing, 2026) show that Gaussian splatting converges to incorrect geometry on glass and mirror scenes, and that spherical harmonics cannot represent the high-frequency, view-dependent color these materials require, even when the method is given perfect ground-truth geometry to start from.
This matters commercially because non-Lambertian materials describe much of what retail and advertising want to present in 3D: eyewear, jewelry, watches, automotive finishes, and glossy packaging.
Miris is delivery infrastructure, not a representation. The method used to capture and encode an asset (Gaussian splatting, mesh, or a future primitive better suited to reflective materials) is chosen by the asset owner. Miris conditions that asset once upstream, then streams it adaptively to any device.
That separation is the point. Non-Lambertian materials are an active research problem, and the primitives used to represent them will keep improving. Because Miris delivers the asset rather than dictating its format, those improvements flow through the same pipeline without changing how content is streamed or embedded. You own the asset. Miris owns the method of delivery.
Streaming video reshaped media consumption; we're doing the same for 3D. Join a small team solving tough spatial streaming, content delivery, and developer experience challenges.
Technical deep-dives on adaptive spatial streaming, the infrastructure challenges of 3D at scale, and what we're learning as we build. Written by the team doing the work.
