Adobe’s Firefly 3 image generation model is here with several new features

One of the most notable new features is auto-stylisation powered by a custom-style engine. This allows Firefly 3 to automatically apply a wide variety of styles, colors, backgrounds, and poses to your AI-generated imagery. It gives creatives much more control and personalisation over the look and feel.

The Structure Reference and Style Reference tools are also getting a big boost. Structure Reference lets you generate new images that perfectly match the layout and composition of a reference image, avoiding the trial-and-error of trying to craft the ideal text prompt. Style Reference then applies your desired artistic style with higher fidelity.

Combining these two features allows you to essentially remix existing images into entirely new artistic compositions based on the structure and aesthetic you specify. Adobe says this enables much faster creative ideation and exploration.

Can you tell if it’s AI generated? (Image: Adobe)

Of course, improved image quality is a central focus. Firefly 3 renders human features, expressions, crowds, and complex structures with vastly more realism. It also better understands conversational text prompts to more accurately depict detailed scenes. Even text rendering quality within the images has been enhanced.

Illustrations, icons, and line art are getting refinements too. The new model produces a much wider range of illustration styles for things like logos, graphics, and concept art.

Firefly Image 3 is launching today in beta via Adobe’s web app, with integration coming soon to apps like Photoshop, Express, Illustrator, and InDesign. Both free and paid plans are available that include usage credits for AI generation.

The company emphasises that Firefly was trained exclusively on licensed stock content to avoid intellectual property issues. So feel free to use those AI-made assets for commercial projects.