The history of artificial intelligence applied to photo editing is enriched with a new chapter. DragGAN is a tool with which to manipulate images in a simple and realistic way. The demo promises miracles.
For some the name DragGAN may not say anything, for others it is already synonymous with “revolution of photo editing”. This is the most recent novelty in terms of Generative AI and while it’s still being developed and perfected, it’s already managed to attract the attention of hordes of enthusiasts and industry professionals.
The why so much success (even on social networks, such as Twitter) is not surprising at all. The creators of DragGAN took care of making things clear, a team that boasts, among others, a researcher from MIT and an expert in mixed reality by Google. In short, not exactly the latest arrivals. The PDF (which you can consult for free here) shows the idea behind this new, revolutionary product editing tool: the simplicity of a gesture known to all (the “drag”) for a realistic and believable result.
Dal white paper we learn that DragGAN allows users to literally drag and drop the content of images generated by GAN (Generative Adversarial Networks). The name is very self-explanatory. And you don’t need who knows what experience with Adobe Photoshop and similar: just indicate a handle point (red) and a target point (blue) to then be dragged with the cursor. DragGAN will take care of all the rest, almost as if by magic.
Il video shared on Twitter by @_akhaliq has already made the rounds on the web. It’s a demonstration of what this model is capable of, so on animal photographs what about people (but also panoramas, vehicles, objects and so on). Starting from a simple click, with DragGAN you can really do everything: open the closed eyes of a person, make an animal completely change its pose, lengthen a skirt, completely change the size of a car and much more. In these pages you will find a series of concrete examples that show “the magic” of DragGAN.
Enthusiasm for DragGAN is skyrocketing, but be careful not to confuse it with other tools Photo editing that someone could (wrongly, in fact) approach him. In fact, unlike Midjourney (just to cite a well-known example), DragGAN doesn’t generate images from scratch, it modifies them. And it is exceptional in this, at least according to the material disseminated on the net up to now.
As anticipated, DragGAN is a tool still far from complete and that’s why it is not accessible to the public. We have to be patient but the premises, it must be said, are excellent.