AI and what it means for Adobe (and Me)

Adobe Photoshop has been using AI for awhile now. Adobe Sensei enabled tools that made it much quicker and easier to select (cut out or isolate) subjects/objects. It also changed the game with Content Aware Fill, a tool that analyzed your photo and used it fill in selections with information taken from your photo. In other words, if you had a photo of your dog but there was an ugly watering can sitting in the shot, you could draw a rough selection around it, then ask Photoshop to guess what might hide the offending object. It would “borrow” pixels from other places in the photo and apply it to the area to make it look as if the watering can were never there. Sometimes it did a fantastic job; sometimes it did not. That’s the nature of AI, and even with bonkers new innovations in photo editing, that factor has not changed.

Recently, a new beta version of Photoshop was released, and it is pretty next level. In fact, it debuts likely the biggest innovation in photo processing software in decades. Using their AI art platform, Adobe Firefly, the beta includes a new feature called “Generative Fill.” Generative Fill is something you can use to remove things from your photo, like Content Aware Fill. The difference is that instead of borrowing information from your photo, it’s using Adobe Firefly to fill the selection. (Adobe Firefly, in turn, uses the art and photography of Adobe Stock to build its images.) I’ve used it for this purpose and more often than not it works brilliantly.

Generative Fill, however, can also be used to generate images, whole cloth. Say you’ve got a photo of a beautiful mountain range, but you think the sky is “bleh.” You can draw a rough selection in the sky and type “fluffy clouds” into the Generative Fill dialogue box. And the–witchcraft? alien technology?– will make clouds appear in your selection. Generative Fill even attempts to match the lighting in your picture. Similarly, you could draw a rough outline around your subject’s clothes and type in “steampunk dress,” and it will change your model’s clothes–just like that. Or will it?

The answer is, “Sometimes.” Generative Fill does really well when it’s working almost like the new Remove tool, that is you’re using it to clean up a photo. It also does well when you want to do things like expand your canvas. Say you want to give your subject a little more room to breathe. You can take a selection tool, overlap it ever so slightly on the area you’d like to expand, and Generative Fill will create more space around your focal point, using the rest of the canvas as a guide. It’s like the Content Aware tool on steroids.

So with all these amazing capabilities at your fingertips, what’s to stop anyone from making loads of point and click art? Well, lots, actually. I tried using Generative Fill to help me save some steps several times and each time the results were iffy to terrible. These were just finishing touches. Simple things like a few blowing leaves or a single butterfly. For one thing, there is a HUGE quality difference between Firefly images and images supported by Firefly engine in Photoshop. I don’t know why there is, but it is a big issue. Secondly, even Firefly has its limitations. Like any other AI-based art generator, it has trouble doing things like rendering hands or understanding certain commands. For instance, I tried to create an image of a person with an eye patch and while it added things *around* the subjects’ eyes, it never actually successfully rendered an eye patch.

And even when Firefly or Photoshop does a good job of rendering something, it’s going to be the result of the AI’s vision, not yours. So if you’re someone who values having complete control of the creative process, if you’re someone who values your own aesthetic, using these new features may not work for you.

The bottom line is that while Generative Fill is a huge game-changer, it is not a job-taker. It’s nowhere near that now. I imagine it will be hugely beneficial to photographers, allowing them to post-process in ways that will speed up their workflow and enhance their creativity. For people like me, who do photo compositing (combined with some digital painting), it will allow me to concentrate more on the fun parts of my work, as opposed to the grunt work. It may help me add some finishing flourishes. But it will not replace me, happily.