2 minute read
Morocco Photo Tour
August 29 - September 10, 2023
porate into changes in the programing to make the AI experience better. Over time, I know it will improve significantly. From my point of view, though, at this time it’s fun to play with but it’s just not ready to be a serious contender in the AI world. I wrote an article last March in Photo Insights about Midjourney’s AI capability, and I love the creative possibilities with this software. That’s why I’m teaching a course on it starting June 10. At this time, Midjourney is light years ahead of Photoshop’s AI capability.
Expanding the canvas with content
Having said this, one great aspect of Generative Fill in Photoshop is its ability to intelligently expand the canvas. AI looks at the image and adds to it in a way that can only be described as brilliant.
For example, the two stallions above were origi- nally shot vertically. I didn’t like how the gray horse was cropped, so I used the pulldown menu command, Image > canvas size. This allowed me to expand the ‘canvas’, i.e. the picture area, to the right. The color that fills the expanded area comes from the background color box at the bottom of the tools palette. Here, it was white. The color doesn’t matter, though, because it gets covered up.
In the image above, I selected the expanded white area as well as a few pixels into the picture on the right side. Enlarge this page on your device and you’ll see the marching ants encroaching on the image.
When a selection is made in the Photoshop beta version, you’ll see an option bar, above, appear at the bottom of your image. Click ‘Generative Fill’, and when that has been done, another option becomes available which says ‘Generate’. Click that, and Photoshop’s AI capability fills in the expanded area based on the original image. You can see the result of this incredible technology below. The tail isn’t quite perfect, but using the clone tool I can fill in the area near the bottom where we can see some of the green vegetation through the hair.
The key to making this work is when you make a selection of the expanded canvas, make sure the selection includes about 30 pixels of the image itself. This tells Photoshop what to do. It may take you a few tries; in other words, if you grab too much of the image, the results may be distorted and unacceptable. If you include too few pixels, there will be an unwanted demarcation line between the image and the new AI portion.
A new and intelligent content aware
When the content aware feature was introduced to Photoshop a few years ago, it was a big deal -- a game changer for many situations in which an unwanted element had to be eliminated. The next generation AI approach is even better. A lot better.
To show you what it can do, look at the before and after versions of a street scene in Ghent, Belgium, on the next page. I thought it would be impossible to get rid of the blurred pedestrians on the sidewalk, but the Generative Fill command did a brilliant job -- although it’s not 100% perfect. I’d still need to do some cloning in Photoshop to restore the cobblestone sidewalk, but look how the software recreated imagery that the pedestrians were blocking from the viewpoint of the camera. This is revolutionary.
To make this happen, click the Generative Fill command, and then choose ‘generate’ from the same mini tool bar. §