Adobe unveils Firefly Image 2 automatic generative artificial intelligence model
At the recently convened MAX 2023, Adobe unveiled the Firefly Image 2, an automated generative artificial intelligence model. Simultaneously, they introduced the Firefly Vector Model, capable of auto-generating vector graphics, and the Firefly Design Model, designed for the automatic generation of design templates. This innovative approach allows creators to expedite the production of high-caliber content via text-generation methods.
Furthermore, Adobe accentuated its commitment to ensuring the security of its auto-generative AI, particularly for commercial applications. The newly launched Firefly Image 2, predicated on the original Firefly imaging model, enhances creative control and performance quality. It incorporates the Generative Match design, which facilitates customized content generation based on user-specified styles. Users can adjust images in accordance with photographic styles via “Photo Settings” or enhance image quality through “Prompt Guidance,” which allows command expansion or rewriting.
Within the architecture of the Firefly Image 2, refinements have been made to skin, hair, eyes, hands, and bodily structures, thereby augmenting portrait and bodily presentation. Additionally, it offers superior color rendition and an enhanced dynamic range. The model leverages authorized content for training data, assets from expired copyrights, and commands in over 100 languages. Paired with a novel paid plan for “Generative Credits,” it significantly reduces image generation time.
The concurrently released Firefly Vector Model specializes in auto-generating vector graphics, producing “Human Quality” vectors. It can generate gradient images congruent with artistic styles through textual commands. Seamlessly integrated with Illustrator for testing, these generated images are editable, can be categorized into groups, and are reusable.
Similarly, templates generated by the Firefly Design Model can be created through text commands, followed by subsequent editing to better cater to users’ specific design needs, thus diminishing the time traditionally required for design.
Currently, NVIDIA has forged a collaboration with Adobe on its Omniverse platform, focusing on OpenUSD (Universal Scene Description). They plan to simplify workflows for developers and designers using Firefly AI technology. Mattel, the world-renowned toy conglomerate known for iconic brands like Barbie and Hot Wheels and recognized as the second-largest global toy merchant, has also streamlined its toy product design process in alliance with Adobe.
