Adobe launched Firefly Image 3 to advance its image-generation services
Adobe recently launched the Firefly Image 3 beta, the latest iteration of its image generation model. Firefly Image 3 brings several advancements to Adobe's image generation solution and powers new features including Structure Reference, Style Reference and Generative Expand.
Firefly Image 3 (in beta), launched at the MAX London Creativity Conference, is available starting this Tuesday on Photoshop and https://firefly.adobe.com. The model advances quality and user control in the image generation experiences by providing higher-quality generated images, displaying a better understanding of prompts—which enables more detailed generations that accurately reflect longer, more complex prompts—, and performing better at text renderings. Overall, Firefly Image 3 supplies photorealistic quality with improved lighting, positioning, attention to detail, and more.
Firefly Image 3 will also power features in the Firefly web application, including Structure Reference and Style Reference in the Text to Image module and the Generative Expand feature in the Generative Fill module, which enables users to change the aspect ratio of existing pictures. The Structure Reference and Style Reference features allow the generation of new images following the structure or style of a user-provided reference image. In particular, the Style Reference feature benefits from the broader range of illustration styles managed by Firefly Image 3. Firefly Image 3 was trained on licensed content, including Adobe Stock, to deliver these new features. Moreover, the model was designed to generate content appropriate for commercial use that does not violate copyright or other intellectual property rights.
The Firefly web application offers free and paid access plans to the tools powered by the Firefly Image 3 beta. The free tier includes 25 monthly Generative credits.
