Back in May, Photoshop Beta received a completely new feature in the form of AI Generative Fill. If you ask me, it was about time the world's most popular photo editing software gave us a taste of what AI can do in its environment.
While Generative Fill is very fun, using it to expand your images takes more time than it should. First, you have to expand the canvas around your image and then select every part of the image where the canvas is empty. If you expanded the image in every direction, that means you now have to select four areas and apply Generative Fill. That's where the new Generative Expand feature comes in!
Before we dig in, I should let you know that you need an active Adobe Creative Cloud subscription in order to have access to Photoshop Beta, which is the only way to get access to the latest AI features in Photoshop (for now).
How To Use Generative Expand in Photoshop Beta to Expand Images?
Adobe describes this new feature as "a new workflow empowering users to expand images with magical results in 1-click in the Photoshop beta.” It may take a bit more than one click to do it, but let me tell you straight away - it's quick and does the job really well.
Let's take a look at an example and learn how to use Generative Expand. I'll use a photo from my recent trip to the magical country of Iceland. About five hours north of the capital of Reykjavik is where you'll find Dynjandi, the beautiful waterfall pictured below.
Dynandi is incredibly large and impressive in person and is definitely one of my favorite waterfalls in Iceland. I took this with my Sony A7IV and Sony 20-70mm f/4 lens and while it's a pretty decent shot (if I may say so myself), I wish I also made a landscape version of the photo. This is where Photoshop's new AI feature Generative Expand comes in!
Let's open Photoshop Beta and open the image. The first thing you have to do is select the Crop Tool from the sidebar (make sure that Generative Expand is selected in the top menu but it should be on by default).
The Crop Tool will allow you to adjust the canvas. Expand it as you see fit by dragging the side handles to the desired dimensions. In my case, I'll expand the image to the left and right. I'll even expand the bottom part just a little bit.
You'll see a tiny window pop-up, which allows you to add a text prompt or leave it blank. We'll leave it blank (in most cases) and I'll explain why later. All that's left to do now is click on Generate and wait a bit for results.
And here's the result. Wow! You have to agree that this is extremely impressive. Although I visited this waterfall recently, I couldn't really tell you this looks exactly like the real surroundings but it's definitely not too far off. Anyway, after clicking Generate you'll be presented with three choices and you can always click Generate again if you're not happy with current results.
So what's the main drawback of this new tool? Can I just print and frame this photo now? Well, not really. Let's zoom the photo to 100% and see what the problem is.
Ah yes, the resolution. You can see the exact line where the real image ends and the low-resolution generative fill appears.
Unfortunately, at least at this point in time, the resolution of the AI-generated part of the image can't be changed. That may be less noticeable when your source image isn't that sharp but if you use an image from a professional camera like me or even a high-quality smartphone, you'll definitely see a difference.
That being said, you can easily use this image online, on Instagram and similar platforms that don't require high resolution.
Adding a Prompt When Using Generative Expand
Remember when I said we should leave the prompt field blank before clicking on Generate? Let me explain why now. Let's use the same image in this example as well, but this time I'll expand it just to the right.
And you know what, let's add a deer to the photo. There aren't really any deers in Iceland, but let's do it for the sake of this guide. When the prompt window appears, I'll type in "deer" and click generate.
Yikes! That's not really what I had in mind. Unfortunately, I found that this is often the case when using prompts with Generative Expand.
In my opinion, the best practice would be to expand your image to the desired dimensions and use Generative Expand without adding a prompt.
Then select a small part of your image where you want to add something (say a deer), and then apply Generative Fill. Let's see what type of results we get that way.
Well, that looks quite a bit better, doesn't it? Sure, technically there are no deers in Iceland but other than that, I'm pretty happy with the results.
Final Thoughts
Honestly, Generative Expand is a very cool feature and I'm glad they brought it to Photoshop Beta. It saves quite a bit of time compared to the standard Generative Fill when you want to expand your photo in a certain direction (or all of them).
I don't want to use AI too much when editing my photos, but I'm certain I'll use Generative Expand here and there, especially if we get support for higher resolution in the future!
