This uses ai to remove objects or add them to a photograph or picture. It can also extend a photo and generate the content to fill the new area. This is really useful for reshaping an image without cropping away pixels. For example, you need to make a picture taller, or wider to fix a certain size or aspect ratio. Its great for extending backgrounds when compositing as well. Another really great use for this is filling in the edges when you make a panorama.

Let’s look at how this works. This is a part of a series, where I’m breaking down the Generative Fill into smaller. more focused tutorials.

EXPENDING AN IMAGE WITH GENERATIVE FILL IN PHOTOSHOP

Here is a wide photograph. In fact, its a panorama that I shot in Kaui.

Let’s make it taller for a composition.

Choose the crop tool. (C key)

Drag the crop up and down to expand the canvas. This is also called outcropping.

Press enter to apply the crop.

The challenge with this is that we now have blank canvas where we enlarged the image. Previously, we would paste in content from other photos and use cloning and Content Aware fill.

Let use Generative Fill (There are a couple of caveats and I’ll address those in this tutorial).

Grab the Rectangular marquee tool

Drag inside the original image, go slightly inside to create a slight overlap for Photoshop to work with.

Inverse the selection so you are selecting the blank areas. Cmd+Shift+I Inverses a selection. (Ctrl+Shift+I on Windows).

Click the Generative Fill button in the Task Bar. If you don’t see it, Window>Contextual Task bar. (Photoshop beta only at this time)

Don’t enter any text into the prompt box. When you click Generate without text, Photoshop will blend the image into the existing pixels.

Click Generate

And you will see that pixels are generated by the ai. These pixels have never existed before, they are generated by machine learning in Adobe Sensei.

There are three alternatives to select from if you look at the Properties panel. For a closer look at each thumbnail, click it.

There are two warnings in this. Let’s talk about them.

1 NOT AT THIS LOCATION, ACTUALLY.

This is not how the scene actually appears, so if accuracy is what you seek, look elsewhere. Where the foreground is located, there is a road. (This is only one instance where using a genuine photo and conventional Photoshop techniques are necessary. Some claim that photography is no longer alive. Not at all. Viva la Fotografia!”

The resolution is low, #2.
The resolution for generative fill is 1k as of this writing. 1024 x 1024 pixels.There are various causes behind this, but the main one is the enormous amount of computer power needed, which is still in development. Scaling is a later process.

But I have a simple solution for this.

Try parts at a time rather than expanding in a single pass and applying the 1k on the entire image.

Reduce the size of your fill area selections.

Create each separately, and with a little extra effort, you can match the resolution of your image.

ADDING SKIES AND OTHER ELEMENTS

The other thing we can do it guide the fill.

Select the top part of the image.

Rather than just filling in the top, lets add a text prompt. In the box, type “Dark moody sky”.

And now we are adding some drama. If you watch the short video at the top, I also add a lake to this scene.

You can use this tool for all manner of images

For example, lets expand from a corner in a city. (Step by step is on the video).

And it does a fairly decent job. It’s not perfect, but can be cleaned up very easily with Photoshop’s suite of tools.

I wanted a mossy brick wall, so I selected the left side of the image and entered in mossy brick wall and lookie!

I hope you found this tutorial useful.