I’ve been playing around with Adobe’s generative AI since the first release of Firefly, so I have a pretty good idea of what their tools are capable of. Despite that, I still find myself being amazed by what you can now do in Photoshop. Here’s an example.
Debs was recently the lucky recipient of a shiny new Philips UltraWide curved monitor with a rather interesting 21:9 aspect ratio. What that means is that regular 3:2 or even 16:9 images don’t always look their best when scaled to fit this screen.
With that in mind, she asked me to automagically make one of her favourite 3:2 desktop backgrounds fit her new screen. Now no matter how I tried to crop the original image, it just didn’t have the same impact, so I decided to see what Adobe’s Generative Fill tool could do. Here’s the original file with white space on either side showing what’s essentially ‘missing’ in order to fit the 21:9 ratio.
Below, after one click and a few seconds, is the result of generative fill.
To be honest, I was quite impressed by the result, and if you’re not familiar with this particular spot (Knoll Ridge on Mt Ruapehu), you’d be hard pressed to tell that parts of the extended image weren’t real (as long as you don’t pixel peep).
Now I did say previously that it only took one click, but if I’m honest there was a little bit of trial and error involved. I did in fact make two separate selections for the left and right and it actually took a couple of attempts to get the desired result. That’s because the amount of of the original image you include in your selection can have a big impact on the output.
There’s also the fact that currently, Photoshop generates three options for you to choose from and not all of them were always usable. In one instance the AI decided that a rather shabby looking pine tree would go quite nicely in the upper left corner.
Despite all that, it still only took me a matter of minutes to produce a final output that there is no way I would have been able to create manually. Needless to say, I’m sold on generative AI, certainly in use cases like this.