AI Magic Eraser Tools
Adobe's Latest AI Tools: Firefly Image 2 Model and Project Stardust
AI Magic Eraser Tools. Adobe recently unveiled a suite of new AI tools, building upon the success of its Firefly AI model. This release comes only a month after the public launch of Firefly AI, signaling Adobe's commitment to advancing AI capabilities. Among the new additions are the Firefly Image 2 Model, as well as enhanced text-to-image features integrated into Adobe Illustrator and Express.
At the forefront of these innovations is "Project Stardust," a tool that has drawn significant attention. Despite its whimsical name, Project Stardust offers remarkable functionality, likened to Google's Magic Eraser, albeit with even greater potency than Adobe's existing AI features like Generative Fill and Generative Expand. As depicted in a teaser video released last week, Project Stardust can identify objects within an image, empowering users to manipulate them by dragging or removing them from the frame. Adobe boasts that this tool not only erases objects but also eliminates shadows and replaces missing pixels with realistic AI-generated content.
During last week's Made by Google event, Google introduced features similar to those planned for Magic Eraser. Adobe, on the other hand, is promoting Generative Fill, which empowers users to add new elements to their images. Additionally, Stardust can now identify objects within an image based on text prompts. Adobe's version allows users to resize and recolor photo objects using text-based instructions.
To enhance Generative Fill's capabilities, Adobe has introduced its new Image 2 Model, which produces higher-quality images featuring more realistic human figures. These figures no longer appear overly plastic or distorted, as seen in previous versions. The text-to-image model can now capture finer details like pores and foliage, and it excels in creating depth of field, resulting in more lifelike images for those aiming to create convincing deepfakes.
The AI has also been significantly improved in recognizing prominent cultural landmarks and symbols. Users gain control over various photo settings, including motion blur and field of view, when crafting an image. Image 2 also boasts a Generative Match feature that applies the style of a different image to a newly generated AI image on a larger scale.
While Adobe expects this feature to save users time, they've introduced a measure to address potential copyright concerns. Users will be required to confirm whether they have the rights to use the generative match image. However, the effectiveness of this confirmation process in preventing potential copyright issues remains to be seen, and it will be interesting to observe how Generative Match reproduces both the style and substance of the original image.
AI integration in Adobe Illustrator and Express
Adobe's foray into integrating AI isn't limited to Photoshop and Essentials; they're now extending its capabilities to Illustrator and Express. Users can harness the power of AI by using prompts to generate vectors in Illustrator and templates in Express.
The Firefly Vector model, featured in this expansion, comes equipped with the Generative Match function, enabling users to create editable vector gradients. This feature closely resembles what's already available in Photoshop, but it's now geared towards generating vector graphics and patterns of "human quality." Adobe has even showcased an example of this technology by presenting an image of a graphic t-shirt with an astronaut cat, all generated using the Firefly Vector model. However, the commercial applications of AI-generated content remain uncertain, particularly in light of U.S. officials' statements that AI-generated content cannot be copyrighted.
Adobe Express follows a similar design path, offering generated text and images transformed into editable designs. With a simple prompt like "rainbow aura fashion show," users can generate a series of posters that incorporate full names and taglines linked to fictitious AI-generated individuals and clothing.
It's worth noting that Firefly primarily relies on content from Adobe's Stock library, a strategy akin to what's been seen in other platforms like Shutterstock and, more recently, Getty Images' AI art generator. Adobe has been somewhat secretive about the specifics of its AI model, leaving us in the dark about the extent of its training data compared to other popular AI art generators like DALL-E or Stable Diffusion XL. This mystery has sparked speculation that Adobe's model might be constrained by a smaller dataset in comparison to the terabytes of data used to train other leading AI models.
Adobe prioritizes having a surplus of training data to ensure the legal integrity of using images for which it possesses the rights. However, Adobe has not provided full transparency regarding its compensation strategy for contributors whose content is utilized in AI training. Presently, their approach includes an annual bonus tied to the quantity of images, vectors, or illustrations utilized for AI training. Similarly, popular stock image platforms have not disclosed their compensation plans for contributors.
إرسال تعليق