Photoshop's AI Leap: The Next Phase of Image Generation

Jun 2, 2023
Kirk Clyne & Suzanne Reeves
A surreal illustration of a realistic 3D brain model in neon colours, seemingly exploding from a black box with a black and white image of a human face on the surface of the box, speaking to the idea of the expansion of the creative mind.A surreal illustration of a realistic 3D brain model in neon colours, seemingly exploding from a black box with a black and white image of a human face on the surface of the box, speaking to the idea of the expansion of the creative mind.

Photoshop's AI Leap: The Next Phase of Image Generation

In a game-changing move, Adobe has introduced a new "Generative Fill" feature to its latest Photoshop (Beta).

Driven by Adobe's Firefly engine, the feature works in a similar fashion to "in-painting" in DALLE-E 2, or the Stable Diffusion Photoshop plug-in: you select an area of the image, then enter a text prompt for what you'd like to see there instead.

An image showing a photo of an house cat indoors on the left, and the same cat wearing a leather jacket and cap on the street at night.
No animals were harmed in the generation of this image.

While Photoshop's existing Content-Aware Fill tries to generate new image content from existing pixels in the image, this AI-driven feature generates imagery from whole cloth — letting you radically alter the contents of an image.

Generative Fill can alter image content, add new content, or even extend an image outward — something that Redditor's are going crazy with this week.

Images of Nirvana's 'Nevermind' album cover and the famous meme of a blonde woman shouting at a white cat have been artificially extended to show more context. The baby appears to be in the ocean near a shark, while the women and cat are in a restaurant with other people at the table who are laughing.

It also promises to match the lighting, perspective and style of the image, too — and from what we've seen, it tends to deliver on that front.

But there's a couple of key things that make this new feature really stand out.

For one, it is non-destructive: your generated content is rendered in layers that can be hidden or removed at will, while the original image below remains untouched.

More importantly, however, Adobe may have just solved a problem that's hampering the use of image generation tools on the job: copyright issues.

Most creative directors we've spoken with have told us they're hesitant to use image generation tools in production because they don't want to run afoul of intellectual property laws and get themselves — or a client — into legal trouble.

But Adobe says its Firefly image generator was trained strictly on Adobe Stock images, openly licensed work, and public domain content — meaning the source material is safe to use in commercial work.

This little detail solves a pretty massive hurdle for agencies interested in using AI commercially, and may alone move Adobe to the front of the pack.

That said, you may want to hold off for a few months. Adobe warns not to use the tool for commercial purposes while it's in beta. They say the tool should all be free of copyright issues in the second half of the year, when it's rolled out widely to all users.

Want to see even more eye candy of this new feature in action? Check out Adobe's announcement post for some impressive images and videos.

Whatcha waitin' for?

We always recommend that designers start using AI tools as soon as possible to avoid getting left behind, and now that one of industries biggest players has entered the fray, we strongly recommend you give their new feature try.

Head to the Adobe Firefly website where you can test out Generative Fill right in the browser, or download the new Photoshop (Beta) if you're a CC user.

Bear in mind that the feature often produces unexpected and surreal results — like two-headed horses, or impossible anatomies. As with all image generation tools, it can take a few tries and more than a few prompt edits to get good results.

But with a little practice, you, too, will finally be able to dress up your cat — without incurring their wrath.

Subscribe

A creative leader, designer, teacher and artist with a 20-year career, Kirk Clyne is well-positioned to help designers master new technologies and build exceptional careers in the creative industry. He co-founded Art & Science, a digital design agency that landed on the Globe & Mail’s Growth 400 list, and spent the better part of a decade teaching digital design courses at San Francisco State University and the Academy of Art College.

Suzanne is an experienced entrepreneur and design leader fascinated by AI's transformative potential. She's run her own successful design firm for the past 10 years, built and mentored design teams, and has degrees in Psychology and Fine Art. Passionate about nurturing the next generation of 'design unicorns', she helps teams and individuals leverage strategy, design, and building skills, believing AI tools can exponentially expand our creative potential.