Introducing Waypoint-1: Real-time interactive video diffusion from Overworld

Waypoint-1 is Overworld’s real-time interactive video diffusion model, controllable by mouse and keyboard with low latency. It’s trained via diffusion forcing on 10k hours of game footage and runs through the WorldEngine inference library for streaming frames in real time. The article covers training details, self-forcing, and optimizations that enable ~30 FPS on consumer hardware.

Source

HuggingFace Blog

Ingested

JAN 20, 2026 · 19:10

Editorial score

4.0 / 5