2 minute read
Generative Fill the new
Adobe has done something that surprised me. It shouldn’t be such a surprise, but I certainly wasn’t expecting it. That’s strange since I love AI so much.
Built into Photoshop now -- at least, the Beta version -- is the ability to create AI imagery in conjunctiom with your photography. Wow! This will revolutionize the workflow and provide additional creative ability.
Actually, there are three new things you can do:
1) You can eliminate objects with a higher degree of sophistication than with content aware,
2) You can expand a canvas in ways never before possible, and 3) you can actually add photographic elements to your pictures that emanate from AI technology.
Download the Beta app
First you have to download the Beta version of Photoshop. Go to the Creative Cloud desktop app, and in the left column click ‘All apps’. You will then see a page with a “Get the Photoshop (Beta) app” button. Click that and you can download the needed version of Photoshop.
Also, make sure you have the latest Photoshop version which, at this time, is 24.5. The beta version is 24.6.
The beta version coexists with the normal version of Photoshop. You don’t use it instead of version 24.5.
Adding elements with AI
If you follow my work, you know I am constantly replacing backgrounds to embellish and improve photographs. I’ve done this within Photoshop for more than 30 years. What is different about the current beta app is the use of AI. All of the backgrounds I’ve replaced up to now have always been with my images. The ancient Irish cemetery below, for example, is a shot of gravestones in the 6th century Clonmacnoise Cemetery in Ireland. The storm clouds I photographed in Tennessee.
On the previous page, the original picture of the same cemetery is shown at left. For the image at right, I used Photoshop’s new feature, Generative Fill, and the software replaced the background with an AI image -- in other words, that castle is not a photograph I took. It was generated from the millions of pictures in the AI data bank.
Similar to all of the other AI software I’ve used, images are created by typing into the appropriate field prompts, which are simply words. So, for the image of the castle in the background of the ancient headstones, I wrote: “Add medieval Irish castle to the background”. You can be more descriptive with many more words, but these are the words I used in this example.
Admittedly, it took me several attempts before the results were acceptable to me. I tried other types of subjects as well. I attempted to add elements with AI images behind horses, a cheetah, and rocks in a landscape, and none of those composites looked good. You’re given three variations to choose from, but in each case none of them worked. Either the lighting was wrong (like sunset lighting behind an element taken in diffused light), the AI image was out of proportion (too large for the situation), or the new image just didn’t make sense. The Irish cemetery composite was the first one that I thought worked.
The chameleon image, above, also worked. The prompts I used were: “Add tropical jungle background”. The focus fall-off looks real, but I had to use the clone tool to eliminate a couple of sunny hotspots in the upper left of the background. No big deal.
So, this is an amalgam of my photography and an AI image that I didn’t take with my camera.
I’ve noticed that AI images introduced behind the photographic foreground aren’t as sharp as the original picture that comes out of a camera. In the image above, that’s not a problem. In most images, though, it would not be ideal.
Keep in mind this is a beta version, and it’s Adobe’s first foray into AI. The reason the company is offering the beta version to us is so we can give them feedback which they will then incor-