ViewNeXt world model flies over any scene from one image, tops LPIPS vs 15 methods

janusch_patas · x · 2026-11-22

A new world model called ViewNeXt lets users fly over valleys, through churches, and down streets along any camera path, generating every frame from just one image or a few. It ranks #1 on LPIPS against 15 prior methods.

Original post →

More from Multimodal

Multimodal channel →