An arXiv paper proposes LEGO, a method that generates egocentric video from a single exocentric recording without depth estimation, point-cloud lifting or reprojection. Instead, an LVSM-style transformer view synthesizer is fine-tuned to render the egocentric view directly, resolving cross-view correspondence internally; its probabilistic mapping averages each region over candidate source locations, which preserves structure but softens fine texture. The authors argue this trade-off suits a diffusion generator that restores detail during denoising, and that the concentration of the correspondence distribution also yields a per-region confidence used to mask low-confidence areas and guide early layout-forming steps. They report it consistently outperforms the explicit state-of-the-art pipeline and generalizes to other datasets without retraining.