Adobe Firefly 2

Adobe Debuts New Firefly Generative AI Models and Features


Adobe has revamped its Firefly portfolio of generative AI models and features as the competition in synthetic media creation intensifies. At the Adobe Max conference, the company showcased the new Firefly Vector and Firefly Design models, as well as the upgraded Firefly Image 2 model and its more realistic and complex visual creations.

Firefly Flock

“Since the release of our first Firefly image model in March, we’ve been taken aback by the amazing response from the community, who has demonstrated how useful Firefly is to enhancing experimentation and design and speeding up workflows, without compromising creative control,” Adobe digital media business president David Wadhwani said. “With over 3 billion image generations to date, including more than a billion in the last month alone, the early adoption of Firefly demonstrates the unique power of bringing foundation models into the workflows customers know and love.”

Firefly Image 2 boasts an enhanced text-to-image function, with improvements tied to new features. That includes an AI better at understanding text prompts, including places and objects. Users can even ask for help improving their prompts before submitting them to be turned into images. Those images can be linked through the new Generative Match tool to an image chosen by the user as a baseline style for the image generator to mimic, keeping the output consistent. Meanwhile, the more photorealistic images of Firefly Image 2, including more convincing humans, better colors, and greater depth of field, can all be adjusted in a way similar to taking a real photo.

The new Firefly Vector model pursues similar goals but with a generative AI model specialized for generating high-fidelity vector graphics and illustrations from text. Users can prompt the creation of logos, website assets, patterns, and other vector images. Firefly Vector also incorporates Generative Match for maintaining a shared style among the generated images. The third new model, Firefly Design, steps back from immediate image use and instead creates personalized templates within Adobe Express, ideal for social media, advertising, and other public-facing content. The point is to save creators time while providing more options to achieve their visions versus the first release. Adobe said auto-generated prompts and guidance also help users refine inputs for better results.

“Our rapid, continuous pace of Firefly innovation continues with imaging, design and vector generation advancements – three new models that set new industry standards for output quality and user control,” Adobe digital media chief technology officer Ely Greenfield said. “Firefly continues to offer the world’s best combination of visual capabilities, designed to be commercially safe, enables transformative new creative workflows across Adobe apps.”

Adobe Launches Firefly Generative AI Models for Synthetic Images and Font Design

Adobe Expands Generative AI Toolkit Firefly to Video

Adobe Brings Firefly Generative AI Tools to Photoshop