Reproduce “Waypoint-1.5: A Real-Time Video World Model for Consumer Hardware” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Rajit Rajpal, Shahbuland Matiana, Liew Wei Pyn, Anmol Agarwal, Ryan Craig, Andrew Lapp, Mithun Hunsur, Sami BuGhanem and 7 more
Waypoint-1.5 makes video generation interactive by predicting each next visual state from recent latents and synchronized keyboard and mouse controls. Its self-forcing post-training and cached representations reduce inference from about 120 model evaluations per frame to four. On an RTX 3060, 360P INT8 reaches 81.88 latent FPS, exceeding the 30-FPS threshold, but this does not measure latency or control accuracy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.