Gauzilla Pro turns smartphone video and drone footage into navigable, photorealistic 3D scenes that run entirely in a browser. No plugins, no installs. The platform handles everything from construction site documentation to retail space visualization, with AI-powered segmentation that lets you isolate complex geometry like rebar, HVAC systems, or MEP infrastructure on the fly. The 4D time-lapse feature is where it gets genuinely useful for AEC teams tracking as-built progress across weeks or months.
Yoshiharu "Yoshi" Sato's built the engine in Rust which compiles to WebAssembly, sidestepping the usual CUDA and PyTorch dependencies that make Gaussian Splatting a hassle to deploy. WebGL handles rendering while WebGPU powers the edge AI inferencing. The architecture uses lock-free multithreading to work around WASM's concurrency constraints. That's a fun surprise given Sato's background: a Master's in Quantitative Finance, years consulting for quant hedge funds on deep reinforcement learning, and before all that, building a 3D graphics engine from scratch in the game industry. That last bit shows.
The core renderer is MIT-licensed on GitHub if you want to understand how EWA splatting math translates to WebGL geometry instancing. The README includes actual equations. For the full platform with segmentation, annotation, and 4D capture, Gauzilla Pro is where the tooling comes together.
