NVIDIA open-source world model breakthrough
Key Questions
What is NVIDIA's SANA-WM world model?
SANA-WM is a 2.6B-parameter open-source model that generates minute-scale 720p controllable video from a single image and camera path on one GPU.
How does SANA-WM advance long-form video?
Its architecture solves key challenges in extended video generation, enabling high-resolution output without heavy compute resources.
What input does the model require?
It takes one image plus a camera trajectory to synthesize realistic, controllable minute-long 720p video sequences.
Why is SANA-WM considered a breakthrough?
It provides accessible open-source world modeling that supports practical pipelines for long-form AI video creation.
Where can developers access SANA-WM?
It is released openly, allowing single-GPU local runs and integration into existing video generation workflows.
SANA-WM (2.6B params) enables minute-scale 720p controllable video on single GPU, advancing long-form pipelines.