Golden hour Japanese temple cinematic scene in Three.js

GOAL: build a 3d environment in threejs and render a smooth, sharp 1 minute cinematic video of it. MOTION: 10 good spots, 7 seconds each, and every shot starts close up and then pulls back to reveal the scene golden hour please. make it look as real as possible and really focus on the light and shadows. match the reference images as close as you can, try to iterate until you reach 99.9% everything has to be code, no assets. the only exception is the background music: find a free cinematic track online that you didn't make yourself and use it put real effort into the details, 2x more than you think is enough. that includes the floor and the ground, it has to look good from above too. light and shadows have to be clean with zero flickering verify and then iterate. keep improving until it's really that close to the references and don't stop before, even if it takes days