Adobe has unveiled new generations of cutting-edge AI image editing tools for Creative Cloud, powered by its new Firefly Image 2 Model.
Introduced at the latest Adobe Max conference, the newly refined Firefly Image 2 Model is an evolved form of its 5-month-old predecessor, Firefly.
The model can create more life-like images by highlighting intricate details such as human skin pores or the subtle play of light on foliage.
Abode is also integrating AI into Adobe Illustrator and Express – users can now prompt the creation of vectors in Illustrator and templates in Express. The Firefly Vector model enables the creation of “human quality” vector graphics and patterns.
Adobe has also refined Generative Fill, ensuring its new model produces higher-quality, life-like images.
Firefly 2 can also effectively manage depth of field, making the resultant images appear even more convincing.
Moreover, the AI is now adept at identifying significant cultural symbols and landmarks. Users can fine-tune photo settings, from motion blur to the field of view.
Adobe says they’ve integrated a total of some 100 new AI features into their software suite.
One highlight in Adobe’s new suite of AI offerings is “Project Stardust,” which parallels Google’s Magic Eraser that detects unwanted elements in photos, like photobombers or power lines, so users can remove them seamlessly.
Abode calls Project Stardust an “object-aware editing engine, ” enabling people to edit images intelligently with little experience.
Project Stardust can detect and manipulate objects in photos, even going as far as removing shadows and replacing the void with realistic AI-generated pixels.
Furthermore, Stardust will enable users to identify objects in images using text prompts. The software also enables the resizing and recoloring of objects in photos, again facilitated by text prompts.
Adobe guns for transparency
Adobe has repeatedly touted that its models are trained on Adobe’s Stock library and not other people’s copyrighted works.
While the extent of Adobe’s AI model training data remains undisclosed, it’s speculated that it might be constrained by the limited available images, especially when compared to the vast data volumes powering other popular AI models.
This might eventually encumber them if they keep upgrading their tools.
Additionally, Adobe has championed the cause of transparency. The company has rolled out an “icon of transparency,” which will be visible on content crafted with AI.
This icon is expected to attach to images created with all major Adobe platforms and there are already talks of integrating it into other services like Microsoft’s Bing Image Generator.
Adobe developed the symbol in collaboration with the Coalition for Content Provenance and Authenticity (C2PA), seeking to create technical standards for verifying the source and provenance of AI-generated content through watermarks and metadata.