All you have to do is mark a section of an image and instruct the program about your desired visual content for that area.
And you can get a result looking like this going from this:
to this:
Photoshop users now have access to a new tool powered by artificial intelligence (A.I.). Recently, Adobe integrated the A.I. capabilities of its recently launched Firefly software into the photo-editing program. This update enables users to make edits to their images using simple text prompts. The software is currently in beta release.
In a promotional video, a cyclist on a country road is featured as an example. Users can select the center of the road and type “yellow road lines” in a text box to remove the lane dividers. If they want to widen the image, they can highlight a box on either side and utilize the “generative fill” option to create a more expansive landscape.
Generative Fill has the capability to enlarge an image’s dimensions, such as transforming a portrait into a landscape. It can fill in the empty areas with fresh content. Additionally, it can eliminate individuals from a crowded city square or generate imaginative landscapes featuring snow-capped mountains, vibrant forests, or mythical creatures.
According to Adobe, Generative Fill ensures that its results appear natural and seamlessly blend with the existing elements in your images by automatically matching perspective, lighting, and style.
In contrast to Photoshop’s Content Aware Fill tool, Generative Fill allows you to precisely specify the additions, removals, or adjustments you desire, instead of relying on Photoshop to make decisions on your behalf. This means you gain greater creative control over how your images are manipulated, opening up more possibilities for customization.
Got a photo of a tiger in the wilderness, but do you want it to appear in the zoo? Just select the tiger, reverse selection (background should be selected), write the prompt “zoo” and click on the “Generate” button. That is it. Here are the before and after photos:
Adobe highlights that the software has been trained exclusively on licensed images from Adobe Stock, ensuring that it respects intellectual property rights when generating content based on text prompts. Many artists use Photoshop, whether openly or not.
Is there a potential for misuse?
A photograph demonstrates the capabilities of Adobe Photoshop’s Generative Fill tool, showcasing a mudflat with mountains in the background. The manipulated version features a car and pond added to the foreground, along with a cloud shaped like a car in the sky.
Text-based AI tools are gaining popularity, and Photoshop’s latest feature is exciting for creatives and photographers. However, there is a concern that it could make it easier to create and circulate fake images that can go viral for the wrong reasons, similar to the manipulated images of the Pope wearing a puffer jacket or Donald Trump in handcuffs.
To address this, Adobe plans to automatically add a metatag called a Content Credential to AI-generated images, labeling them as such. The intention is to prevent the misuse of images created with Generative Fill by distinguishing them from genuine photos. However, if people can remove or manipulate this metadata before sharing an image, it may not effectively counter AI-generated fake news.
The extent of these potential issues remains uncertain. What is clear is that AI is revolutionizing our apps and transforming how we interact with technology, and Photoshop’s Generative Fill represents an impressive example of this advancement.
Adobe is actively promoting Content Credentials as a solution to combat the proliferation of fake images. Back in November 2019, Adobe announced a collaboration with Twitter and the New York Times Co. to establish the Content Authenticity Initiative. The main objective of this initiative is to create a widely accepted industry standard for attributing digital content.
Since then, Adobe has developed a feature called Content Credentials, which it describes as “nutrition labels” attached to images. Each label aims to provide an unalterable and secure record of an image’s origin and its history of modifications. This feature is currently still in development.
