According to the researchers, the true significance of this work lies not only in the image manipulation itself but also in the user interface. While AI tools like Generative Adversarial Networks (GANs) have been able to generate realistic images for some time, most methods lack flexibility and precision. In previous approaches, requesting an AI image generator to “create a picture of a lion stalking through the savannah” would yield a result, but it might not precisely match the desired pose or requirements.
The DragGAN model offers a promising solution to this limitation. The user interface closely resembles traditional image-warping techniques, but instead of merely distorting existing pixels, the model generates the subject anew. As the researchers explain, their approach can even “hallucinate occluded content, like the teeth inside a lion’s mouth, and can deform following the object’s rigidity, like the bending of a horse leg.”
