Adobe Photoshop Firefly Generative AI is so good, it’s addictive.
This is not supposed to be the case with a powerful productivity tool like Adobe Photoshop. Preferably, you open the Mac or Windows app, do the job of editing your photos, close it, and away you go. It’s not a game, it’s not fun. actually?
With the new Generative Fill tool, Photoshop is ridiculously fun and powerful. I have now spent hours completing or letting the AI complete all kinds of images.
Released late last month, this new Photoshop beta looks a lot like previous versions of Creative Cloud, but once you start messing around with themes and backgrounds, Firefly-powered new generative fill tools step in to help you completely transform your photos with a single click. Just.
Having used Photoshop for decades, I realized my way around the app and thought a little about how to use these new tools, but it wasn’t until I saw a TikTok where someone perfected a popular meme using Photoshop Beta Generative AI that I knew I had to try it myself.
I started by finding about a dozen classic memes (Leonardo DiCaprio lopeAnd Keanu sadAnd success childAnd Her boyfriend is distractedAnd Disastrous girl) And then, through some experimentation, they figured out how to expand the frame to imagine the rest of these iconic and often used images.
In order to imagine what would happen, say, going around the distracted friend, I had to expand the canvas of the image and then reposition the original so that there was ample white space (the background).
Once I create a little white space around the original image, I can select it with the magic selector and then select the generative fill button that appears automatically.
Doing so launched a prompt box where I could describe how I wanted the Firefly-powered AI to fill the picture. I found, however, that the system usually worked better when I left that space blank and hit Create.
Generating AI images usually takes about 15 seconds on a MacBook Air M2. It should be noted that Adobe Photoshop Generative AI is based on a cloud-based system, which means that it will only work if you are connected to the Internet.
In the case of Ermahgerd Girl (a portrait of a very excited girl clutching her favorite Goose Bumps books), the results were shockingly good. Adobe Generative AI completed the background and made a complete body and costume set out of little more than parts of a white T-shirt and jacket.
I actually ran the Generative AI three times and got different outfits and backgrounds. They were all very convincing and not just because AI hit fashion aesthetics. The resulting image followed the overall image quality and color tone and blended together seamlessly.
It’s not perfect. The girl’s fingers seem blended together and I can’t tell if one of her hands has five or seven fingers.
In the distracted friend program, Adobe Generative AI did a better job of using fingers and creating a more convincing overall image, with a few exceptions. There are some small gaps between the original image and where it was created, and the legs of two figures may have been entwined in one crooked, jean-covered limb.
Some of the generative images were disguised and comical. The resulting shredded Shibu dog body turned into a throw covering the sofa.
Adobe Generative AI is almost perfect Scumbag StevePants, a fur jacket, and a hoodie worked for the final entrance, which we haven’t seen before.
There were some really bizarre results like the disaster girl image that Adobe Generative AI turned into a complete movie-worthy disaster.
For Sad Keanu, I started with a fairly tight set of the actor on the bench and let Adobe Generative AI finish the bench and the rest of the actor’s body. At a glance, the image looks solid and exactly the same as the full, uncropped image. However, closer examination reveals that the seat is poorly constructed and Keanu appears to be sitting on air. Also, his fingers are a bit sausage-like. Ai managed to put what looked like a burrito in his hand, but there was no discarded lunch on the bench next to him.
Later, I experimented with Adobe Generative AI on some of my own images. The results were often very good, although it works best when the subject is closer to the center of the original frame.
Sharper images also work better. The original Real Housewives photo is a bit blurry, mostly because people are engaged in a shouting match. The first attempt at generative AI to complete the picture failed miserably. On a second try, I wrote “complete women”. The result was comically insane.
Finally, I used the AI selection tool to remove the subjects and essentially remove some of the images. Girl disaster, for example, strikes differently when you take away the little girl.
Overall, the results ranged from realistic with barely noticeable glitches and antics to fun artistic and sometimes a little plastic. Some look perfectly reasonable at a glance but, upon closer inspection, look hideous.
If you think these results aren’t good enough to use, just wait. This is only the first beta and Firefly and Adobe Photoshop Generative AI are guaranteed to improve. By the time this app goes from beta to final version, we may not be able to tell the difference between the original memes and those that have been fleshed out with a single mouse click.
“Lifelong beer expert. General travel enthusiast. Social media buff. Zombie maven. Communicator.”