Photoshop AI: Generative AI review

In March 2023, Adobe announced the release of Firefly, its AI image generator. After being in public beta and available through Photoshop Beta for months, Adobe has now officially launched Firefly to the public. With this release, all AI tools are now available through Photoshop. In this review of Photoshop AI, I will cover everything from how to get and use these AI tools to my experience and opinions.

What is the Photoshop AI?

Photoshop AI is the collective term for all AI tools available through Photoshop, including widely talked-about Generative Expand and Generative Fill to more mainstream tools like Content-Aware Fill, Object Selection, Remove, Match Fonts, and Sky Replacement.

Photoshop AI provides a significant advantage as it can perform complex tasks in just a few seconds. With traditional image editing, adding a subject to your images requires a lot of experience and finesse. However, with AI tools, most of the hard work is done for you, making image editing more accessible to the average user and reducing the need for years of experience. This advancement also makes life easier for professionals.

How to get Generative AI in Photoshop

To access the powerful capabilities of Generative AI, you will need an active subscription to Photoshop, which can be obtained through one of three available plans. If you’re not yet a subscriber, you can test Generative AI for free with a 7-day trial. Additionally, Adobe Express offers another convenient way to access Generative AI, whether online or on a desktop.

If you’re an existing Photoshop customer, all you need to do is update Photoshop to the latest version (Photoshop v25.0). The update reads, “This update contains new Adobe Firefly powered features: Generative Fill & Generative Expand and update to the Remove Tool. View more.” Here’s how you do it:

• Open the Creative Cloud app
• Select the Updates tab in the left sidebar
• Find Photoshop and click Update

Generative Expand

Ever found yourself wishing you could extend the beautiful sunset sky in your beach photo without it looking like a hasty copy-paste job? Generative Expand is the feature that fulfills that wish. It’s integrated as part of the Crop tool workflow, so it blends seamlessly into your editing process. You simply expand your canvas and let Generative Expand fill in the gaps—literally. It can either match what’s already there or inject something new based on your text prompt. It’s your world; Adobe’s just helping you paint it.

Both Generative Fill and Generative Expand are trained on Adobe Stock’s colossal database of high-res, high-quality images. So, rest easy knowing your creations are commercial grade and royalty compliant. Even better, Adobe’s initiatives like the Content Authenticity Initiative ensure that your generated images are ethically sourced and transparently edited.

Generative Fill

Ah, Generative Fill. It’s the sort of thing that makes you go, “Why didn’t someone think of this sooner?” Whether you’re looking to add a stray cat to your alleyway snapshot or remove that awkward soda can from your picnic photo, Generative Fill is your go-to. A few text prompts and voila—you’re basically a wizard. And no, this isn’t just pasting one part of the image onto another like you used to do in MS Paint as a kid. This AI is smarter than your average bear. It uses deep learning to match your image’s lighting, perspective, and style. Your results are saved in a Generative layer, so go wild with Photoshop’s classic features afterward to make it even more you.

The user interface is also forgiving. When you’re first onboarded, there are videos and tooltips that guide you through generating your first asset, so you’re never flying blind.

Before you dive in, keep a few things in mind. You need to be online to use Generative Fill, as it relies on cloud processing. Also, both features are designed for commercial use and are as safe as a vault. These aren’t your garden-variety image editors; they are robust tools built on cutting-edge AI, ensuring that you’re only limited by your imagination.

My take on Photoshop AI

Generative Fill and Generative Expand are intriguing tools to play around with, but they currently fall short of the realism needed for commercial work. That’s not to say they’re useless; they excel in the realm of illustrations, where the expectation for hyper-realism is often more lenient. In these contexts, the tools perform well and can deliver some creative and visually pleasing results.

Firefly, Adobe’s underlying AI, performs a bit less impressively compared to some other tools on the market. One reason could be that Adobe has been careful with the volume of images used to train Firefly, focusing on fair use and licensing considerations. This cautious approach potentially limits the tool’s efficacy and realism.

That being said, the future looks promising. As more data is fed into these AI-driven tools and the algorithms behind them continue to improve, it’s reasonable to expect that Generative Fill and Generative Expand will soon be robust enough for commercial applications. In about a year, I anticipate we’ll see a level of realism that truly integrates with professional workflows.

Other Photoshop AI tools

Content-Aware Fill

Ever snapped a perfect photo, only to later notice an annoying photobomber? Enter Adobe Photoshop’s Content-Aware Fill—a feature that’s like autocorrect but for your images. Just select the intruding object, right-click, and choose “Content-Aware Fill.” Voila, the object is replaced with background elements as if it was never there.

The real beauty lies in the details. Content-Aware Fill isn’t just a one-trick pony; it offers a full suite of controls. You can fine-tune your selection, adjust color and brightness, and even rotate the fill pattern for a seamless blend. All of this happens in a dedicated workspace that features a real-time, full-resolution preview.

But the icing on the cake? You can tackle multiple areas without leaving the workspace, a feature introduced in Photoshop 21.1. This makes it a versatile tool, whether you’re editing out minor annoyances or doing more complex graphics work.

Neural Filters

Remember when adjusting a smile in a photo required a deep dive into complex editing? Photoshop’s Neural Filters change the game. It’s like having an artificial intelligence assistant tucked right into your Photoshop workspace. Powered by Adobe Sensei’s machine learning, this feature lets you make significant changes or create fantastical effects with just a few clicks.

The interface is pretty user-friendly, too. Navigate to Filter > Neural Filters, and behold—there’s a buffet of filters, both in the ‘Featured‘ and ‘Beta‘ categories, waiting to be explored. Want to make your subject look older or younger? There’s a filter for that. Dreamy landscapes or sky replacements? Yup, they’ve got you covered. And if you’re feeling adventurous, the ‘Beta‘ tab is where future tech is being tested.

And here’s the kicker: Neural Filters generate entirely new pixels to blend seamlessly with your image, meaning they’re creating elements that didn’t exist in your original photo. You’re not just moving pixels around; you’re essentially co-creating with AI. The output options are flexible too—you can have a new layer, apply changes destructively to the current layer, or even generate a new Photoshop document.

Match Fonts

The Match Fonts feature in Photoshop has become more robust since its improvement in the 21.2 release (June 2020). No longer is it confined to a limited set of fonts; the tool now supports vertical text and multiple-line detection. It’s a perfect solution for those who’ve stared at a piece of text, puzzled over its font, and wondered, “What font is this?” With Match Fonts, Adobe Sensei steps in to remove the guesswork and delivers a list of similar fonts straight to your interface.

Here’s how you can use it:

• Select the Rectangular Marquee Tool and make a selection around the text in your photo.
• Select Type > Match Font. Photoshop displays a list of fonts similar to the text in your selection.
• Find matching font in an image
• In the Match Font dialog, use the options below to filter the results.
• (Optional) Choose a Type Option – Roman or Japanese.
• (Optional) Deselect Show fonts available to activate from Adobe Fonts to hide fonts from Adobe Fonts and only view fonts available locally on your computer.
• In the matching fonts list, click the font that is closest to the font in your photo.

There are some best practices to keep in mind to get the most accurate results. For example, using two to three lines of text for detection typically yields better matches than just a single line. And, it’s essential to correct any skew or perspective issues in the image beforehand to ensure optimal performance. The feature works mainly for Roman/Latin and Japanese characters at this point, so it does have its limitations in the types of text it can successfully match.

Although Match Fonts is a helpful tool, there’s room for growth. It could benefit from expanding its character set support and perhaps even incorporating an option to detect mixed typefaces in a selection. The technology is getting there, and as machine learning algorithms become more sophisticated, the tool will only improve.

FAQs