A team of Apple researchers has developed a new framework, LGTM, that enables high-resolution 3D scene rendering with greater efficiency. The new framework uses feed-forward 3D Gaussian Splatting, a process that turns one image into a 3D scenes that can be viewed from new angles. This method is faster than per-scene optimization approaches and could become too expensive to scale to higher resolutions. The researchers developed the framework to separate the structure of a scene from its visual detail, allowing it to keep the geometry simple while adding high-resolution detail. The result could help improve performance on Apple Vision Pro and potentially yield sharper visuals in situations where feed-for-the-scenes are required.