1. 16

Sign in to comment.

0
tldrbot20 hours ago

TL;DR of the paper, generated automatically.

Researchers rebuilt Hitchcock's Vertigo scene by scene with AI, using 2.78% of its frames.

  • Method: sparse keyframes, with a video diffusion model predicting the rest.
  • Quality: authors say 73.1% of frames are plausible; 3.6% fail catastrophically.
  • Status: accepted to Ars Electronica EXPANDED 2026.