“Starting Veda. The idea: turn a phone video into an explorable 3D scene in the browser using Gaussian splatting. The problem I immediately hit — my computer can’t handle 3D rendering or the CUDA training that normal splatting pipelines need. So before writing real code I spent 90 minutes just researching whether this is even possible without a GPU. Found a possible path: skip the neural training, go straight from Structure-from-Motion sparse points to isotropic Gaussian primitives. Sacrifices quality but might actually run on my hardware. Got the initial project scaffolding down. Doesn’t work yet. Next: get COLMAP extracting camera poses from a test video.
