This project is a walk through capture app with a React frontend and FastAPI backend that turns uploaded recordings and 3D models into step-by-step, spatially anchored walkthroughs. It currently supports uploading media and model files such as .splat, .ply, and .gz, running a pipeline to transcribe recordings and extract steps, displaying the scene in a custom 3D viewer, placing step anchors and free form notes in 3D space, and using AI to suggest spatial labels for anchors.