NVIDIA Announces SANA-WM, an AI That Can Generate 1-Minute Videos and Precisely Control Camera Movement

NVIDIA released SANA-WM, a 2.6B-parameter open-source world model for controlled video generation.

The model can generate up to 60 seconds of 720p video while following detailed 6-DoF camera paths.

It uses hybrid linear attention, a two-stage camera control pipeline, and a refiner to improve consistency and quality.

NVIDIA says the system is efficient enough to produce long videos on a single GPU.