Photoshop’s AI Upgrade: Image Generation, Background Swaps

Adobe has unveiled a substantial upgrade to its renowned graphics editing software, Photoshop, introducing a host of new AI-driven features. Among these enhancements is the groundbreaking ability to generate images from scratch, a first for Photoshop. Users can now start with a blank canvas and simply input a text prompt describing the image they wish to create, empowering them to unleash their creativity with unprecedented ease.

Another notable addition is the option to replace backgrounds with AI-generated content, offering users the flexibility to transform the context of their images effortlessly. Whether it’s swapping a grassy field for a sandy beach or any other desired setting, Photoshop now provides three different background alternatives that adapt to the lighting, size, and positioning of the subject in focus. This feature opens up new possibilities for artistic expression and visual storytelling.

Additionally, Adobe has introduced the ability to supply Photoshop with reference images, allowing users to mimic the style of these images in their creations. This feature is particularly useful for maintaining consistency across designs or for achieving specific visual aesthetics.

Furthermore, Photoshop now offers an option to detect and remove people from photos automatically. This feature can be handy for scenarios such as removing tourists from a landmark photo, streamlining the editing process and enhancing the overall composition of the image.

The AI upgrades extend beyond Photoshop, with Adobe integrating generative AI features into other apps in its portfolio, including Lightroom and the publishing package InDesign. These enhancements are part of an upgrade to Adobe’s AI engine, now dubbed Firefly Image 3, which promises better rendering of lines and structures, as well as an expanded range of image outputs.

While these AI advancements represent significant progress, Adobe acknowledges that there are still challenges to overcome, such as improving the resolution of generated images. Currently, the resolution tops out at around 1,500 x 1,500 pixels, which may limit the use of these images in high-resolution applications. Adobe’s chief technology officer, Ely Greenfield, has outlined potential solutions, including increased compute power, upscaling techniques, and generating images piece by piece.

Despite these challenges, the quality of images generated by Firefly AI has improved dramatically since its inception. Faces, hands, and text, which were previously challenging to render accurately, now exhibit marked improvement. Adobe credits these improvements to better training data and a focus on “prompt adherence,” ensuring that the AI delivers on user requests accurately.

The AI upgrades to Photoshop and Adobe’s Creative Suite apps are being unveiled at Adobe Max in London, signaling a significant leap forward in AI-driven creativity and design capabilities.