Translated data: GPT4Motion integrates the GPT-4 large language model with the Blender physics engine to propose a new text-to-video framework. By utilizing Blender scripts and the physics engine, it efficiently generates coherent and realistic object motion videos. Experiments have demonstrated significant achievements in motion consistency and physical accuracy, covering a variety of materials including rigid objects, cloth, and liquids, bringing new breakthroughs to the field of text-to-video technology.