Down the line, Cube 3D will be able to generate objects using images as inputs, too. “It will ultimately be a multimodal model, trained on text, images, video, and other types of input – and will ...