The dog on the right was not in the original image, but was added with a few instructions in plain English: “Place another French bulldog on the right of the image.”

A new test version of Photoshop, the popular image editing app by Adobe, incorporates generative artificial intelligence tools – and the results are, for the most part, incredible.

TechCentral put the new beta of Photoshop through its paces, testing the new AI tools, which allow users simply to type in commands to have them create or modify an image – for example, selecting an image of the ocean and asking Photoshop to add a fishing boat to the scene works like magic.

Adobe recently made waves in the generative AI field with a series of ground-breaking announcements. Following the beta release of its Adobe Firefly generative AI tools and their integration into Google Bard, Adobe has now incorporated the same technology into its flagship image editor.

The addition means Photoshop users can now effortlessly add, transform or remove elements in their images using simple text prompts.

The key feature, called “generative fill, allows users to select a specific portion of an image using selection tools like the lasso and fill it with new imagery based on a text prompt. The generated content automatically matches the perspective, lighting and style of the original image, even adding details such as shadows and reflections where appropriate.

To ensure flexibility and non-destructive editing, the newly generated content is added as separate layers, allowing users to revert changes without affecting the original image.

Alongside Generative Fill, Adobe has introduced around 30 new adjustment presets – filters users can apply to achieve a specific look and feel they are after.

Remove tool

Additionally, the new remove tool utilises Adobe Sensei AI to eliminate unwanted objects, potentially saving hours of manual work. A contextual taskbar has also been implemented to improve accessibility by recommending relevant next steps in various workflows. Lastly, an enhanced gradients feature offers new on-canvas controls for better gradient manipulation.

These new features are a logical and anticipated progression in Adobe’s expansion of generative AI tools. Adobe has been consistently enhancing Photoshop with AI-driven tools such as neural filters that transform facial features.

With the introduction of generative fill, Photoshop could radically change the efficiency and ease of photo manipulation. By automatically matching the context of an image, the time-consuming tasks of adding details like shadows and reflections are significantly reduced. Moreover, users may no longer need to spend time searching through stock photo libraries to find suitable images for their compositions.

The original image of the Johannesburg skyline

“Add a flock of birds” and “Remove the building”. Note how Photoshop “fixes” the red building at the base of the removed tower

“Add a small fishing boat” and “Add a dark and stormy sky”

Can you spot all four changes, all done using generative fill?

Perhaps most importantly, generative fill promotes experimentation, allowing users to test unconventional ideas. This instant feedback loop encourages users to explore wild concepts and pushes the boundaries of creativity.

The AI model is trained on Adobe Stock images rather than images found online, ensuring reliable and trusted content that can be used in commercial projects. Generative fill supports Adobe’s new “nutrition labels”, which serve as tags to indicate that an image has been created or edited using AI.

Not all the adjustments TechCentral tried worked well. Even so, it’s still early days and overall the AI did an impressive job. Adobe’s integration of Firefly’s generative AI technology into the new Photoshop beta represents a significant milestone. – © 2023 NewsCentral Media