Photoshop Beta Generative Fill

This year, it seems like everyone is talking about the potential of artificial intelligence and intuitive learning. In fact, it's become a common topic among my circle of friends, colleagues, and clients. We're all familiar with the ever-present ChatGPT and numerous emerging AI software that claim to make our lives simpler and expedite our workflow. As someone who works as a graphic designer and business owner, I try to stay up-to-date with the latest technology and software. However, even with all my efforts, I know that I have only scratched the surface of what this technology can do and its true power.

Recently, I was extremely excited and deeply intrigued when Adobe announced that they were integrating artificial intelligence into one of their leading creative software - Photoshop. With the latest release of Photoshop Beta, users will find a new tool – Generative Fill. This new tool is like magic, akin to the Elder Wand. With a simple selection of the marquee or lasso tool, users can now tell Photoshop what they want AI to do in that space. For example, you can remove a person or obstruction from an image or add a table and coffee mug to a space where it never existed before. The limits are truly endless.

Although results from emerging AI software are never perfect, in my experience using generative fill, it's incredibly close to being perfect. Every time you make a selection and give Photoshop the instruction of what to do, you will be presented with three options. You can easily click through the options to see which one works best or regenerate new results with a single click. All in all, it's exciting to think about the endless possibilities that AI software can bring to the table and I can't wait to see what's next.

As a quick example to show what generative fill can do and to learn more myself – I started playing around with an image I pulled from Unsplash.com. First I opened up the image in Photoshop Beta, and immediately wanted to see what kind of edits and alterations I can make without distorting the image or making it look ‘photoshopped’.

‍

‍

‍

‍

I wanted to see how Generative Fill would work when it did not have anything native to base things off of. For instance, removing the sunglasses and creating the mans eyes which we can't see from the original image. I used the Lasso tool to select the sunglasses, and the only prompt I used was “remove”.

‍

‍

‍

‍

‍

‍

Okay, things were off to a strong start here. The mans eyes looked spot on realistic, with the coloring, shading, and even directing his eyes towards his hand where it is pointing. Mind you, this took a grand total of about 6 seconds to edit. I was shocked.

Following that I wanted to change up his clothes and see what kind of results we can get. Again, using the lasso tool I loosely selected his denim jacket, and prompted Generative Fill with “Leather Jacket”. Boom. The shadows are perfect, the lighting is correct, even some of the reflections in the shiny leather correctly identify where the sun is in the image. This is wild!

‍

‍

‍

‍

‍

One of the last things I wanted to play around with on this image is expanding the sides and turning it from a portrait image into a classic 16:9 landscape. Like everything else with Generative Fill, this was incredibly easy. The first step was expanding the canvas size. As you can see, we’re left with a whole lot of empty space along the sides. From here all I had to do was use the marquee tool to select both the left and right sides with just a little overlap of the original image. Open up the Generative Fill prompt box, but this time just leave the prompt blank. This tells the AI to just fill in the blank areas to match the rest of the image.

‍

‍

‍

‍

‍

Just like that Generative Fill builds out the missing parts to the picture in a way that naturally reflects our starting image. It’s incredible how powerful this tool is. In the past, if you wanted to expand an image like this you would have to blend a second image to the left and right sides and and have layers on layers of lighting edits and corrections to make it look realistic. Now, it’s as simple as a few buttons and Generative Fill absolutely nails it.

‍