Stanford & NVIDIA AI makes 60-second videos with full stories from one prompt
A new AI method from Stanford and NVIDIA is generating one-minute-long videos with full scenes, characters, and storytelling. By adding special "test-time training" layers to existing models, researchers created continuous clips like a full Tom & Jerry cartoon—without stitching or editing. The technique could soon change how AI videos are made.
Just a few months ago, AI-generated videos were still mostly limited to short, single-scene clips. Now, a new method by researchers from Stanford, NVIDIA, UC Berkeley, and others is shaking that up—by generating one-minute-long AI videos that actually tell full stories, with multiple scenes, motion, and continuity.
We came across this while watching a generated Tom & Jerry video online. The entire thing played out like a proper cartoon episode—pie theft, a chase, a heartfelt reunion—all in just one AI-generated shot.
Read more:
_1744280316.jpg)
_1749963941.jpg)
_1749964328.jpg)
_1749964731.jpg)