At the Adobe Max conference, Adobe wowed the creative world with the unveiling of “Project Stardust,” an innovative AI tool that promises to revolutionize photo editing. With the catchphrase “Magic Eraser on Steroids,” this cutting-edge tool takes photo manipulation to a new level.
“Project Stardust” is akin to Google’s Magic Eraser but with a plethora of advanced features. It can identify objects in a photo and enable users to manipulate them by simply dragging them around the frame or deleting them. What sets it apart is its ability to seamlessly remove shadows and replace missing pixels with realistic AI-generated content. Moreover, users can employ “Generative Fill” to add new objects to images, making it a versatile tool for content creators.
In addition to object manipulation, “Project Stardust” can identify objects in an image based on a text prompt, allowing users to resize and recolor objects with ease. Adobe’s new “Image 2 Model” enhances the quality of generated images, ensuring that human figures look more realistic and less artificial. This innovation includes improved detail rendering, such as pores and foliage, as well as enhanced depth-of-field generation for strikingly realistic results.
One remarkable feature of “Project Stardust” is its proficiency in recognizing cultural landmarks and symbols. Users can fine-tune various photo settings, including motion blur and field of view, during image creation. The “Generative Match” feature is another highlight, enabling the application of the style of one image to another AI-generated photo at scale. However, Adobe requires users to confirm their rights to use the generative match image to address potential copyright concerns.
AI Integration in Adobe Illustrator and Express
Adobe’s AI integration doesn’t stop at Photoshop or Essentials. The company has extended its AI capabilities to Adobe Illustrator and Express. Users can now generate vectors in Illustrator and templates in Express through simple prompts.
The “Firefly Vector” model, which incorporates the “Generative Match” feature, can create editable vector gradients. This feature, reminiscent of Photoshop, now produces “human-quality” vector graphics and patterns. Adobe showcased an example of an astronaut cat graphic on a t-shirt, highlighting the creative potential of the Firefly Vector model.
Adobe Express offers a similar experience, with the added ability to generate text along with editable designs. A single prompt can generate multiple posters, complete with names and taglines associated with AI-generated people and clothing.
Adobe’s Approach and Legal Considerations
Adobe’s Firefly AI tools rely on content from its Adobe Stock library. While this approach differs from the vast training datasets of other AI generators like DALL-E, it provides legal safeguards since Adobe owns the rights to the images used. However, the full scope of Adobe’s training data remains undisclosed, leading to speculation about potential limitations compared to other AI models.
Regarding compensating contributors, Adobe has disclosed minimal information about its plans, mentioning an annual bonus based on the number of images, vectors, or illustrations used for AI training. This mirrors the secrecy surrounding compensation plans by popular stock image sites. The intersection of AI and intellectual property rights continues to be a complex and evolving issue in the creative industry.