PhysVid: Physics Aware Local Conditioning for Generative Video Models

Saurabh Pathak, Elahe Arani, Mykola Pechenizkiy, Bahram Zonooz · 2026

Read PhysVid: Physics Aware Local Conditioning for Generative Video Models by Saurabh Pathak, Elahe Arani, Mykola Pechenizkiy, Bahram Zonooz (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-26285.txt.

Description

Pathak, Arani, Pechenizkiy, and Zonooz present PhysVid, a way to make generated videos obey basic physics. Video models often look sharp yet break simple physical rules. PhysVid adds local notes on states, contacts, and limits for short chunks of frames, fused with the main prompt. At run time, negative prompts name local law breaches to steer away from them. On VideoPhy, commonsense physics scores rise by about 33% over base models.

Catalog record

CREATORSaurabh Pathak, Elahe Arani, Mykola Pechenizkiy, Bahram Zonooz
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46427
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.26285

Browse the full library · About the Tower of Babel