Two reusable components on one small engine. The layered view is ordinary elements moved by CSS 3D transforms, so it adds no download and its content stays real text. The model view explodes a 3D part with three.js, which is fetched only as the figure nears the screen and draws a frame only when something changes. Both follow scroll, hover, a tap or your own control, and both stand still when reduced motion is on.
Use the layered view when the thing being explained is flat: a page, an app screen, a stack of services. It adds no download and its content stays real text. Use the model view when it is a physical part, and give it a poster image for the moment before the engine arrives.