Microsoft researchers use GANs to generate images and storyboards from captions

In a new paper, Microsoft researchers propose a system -- ObjGAN -- that can generate images from no more than text captions.