Grenzwert loads, and suddenly you're staring into a skull. Not a stylized mesh or a simplified approximation. Actual volumetric CT data, path-traced in your browser with the kind of fidelity that resembles a medical imaging workstation. Mikhail Gorobets built this as a showcase for ground-truth volumetric rendering, and it delivers. The transfer function editor lets you peel back layers of tissue in real time, adjusting opacity and color mappings while the renderer keeps up. You can crop into the volume in 3D, slice away what you don't need, and watch light scatter through bone and soft tissue with surprising physical accuracy.

The technical story here is layered. The core is a cross-platform C++ engine compiled to WebAssembly, with WebGPU handling the heavy lifting on the GPU side. Progressive streaming keeps things responsive: you get a coarse mip level first, then finer detail loads in as you idle. It's the kind of architecture that respects both the network and the user's patience. Path tracing volumetric data is computationally brutal, but the progressive refinement means you're never waiting on a blank screen. You interact, it responds, then sharpens.

Mikhail Gorobets is a graphics engineer whose work sits at the intersection of medical visualization and real-time rendering. The full source is up on GitHub if you want to dig into the shader structure or see how the streaming pipeline works. Worth a look if you've ever wondered what it takes to make WebGPU do something genuinely challenging.