Adobe has integrated its next-generation AI model, Image 3, into Firefly, its platform for generative AI tools. This marks a significant advancement in image generation, offering users greater control and significantly improved image quality.
Firefly’s Text to Image module and the latest public beta of Photoshop (version 25.9) now offer access to Image 3 in beta testing. A key feature of Image 3 is the ability to leverage reference images to guide the generated output, empowering users to achieve a higher degree of precision in their desired visuals.
Adobe’s Firefly Suite: Powered by the Image 3 Generative AI Model
Debuting in 2023, Firefly established itself as a comprehensive suite of generative AI tools for artistic creation. Accessible as both a web application and through integration with Adobe Creative Cloud products, including Photoshop, Firefly has undergone continuous development.
This ongoing evolution is exemplified by the introduction of successive AI models. Following the release of the Image 2 model in October of last year, Adobe has now unveiled the third generation, Image 3.
Improved Quality, Text Prompt Fidelity, and the Introduction of the Structure Reference System
According to Adobe, Image 3 represents a significant leap forward in the quality and detail of generated images. This advancement translates to a higher degree of fidelity with which the model adheres to users’ text prompts.
Specifically, the new AI model demonstrates enhanced capabilities in generating realistic images of people, including crowds. Additionally, it excels at producing plausible text within illustrations and icons.
Furthermore, Image 3 offers expanded options for tailoring AI output. One such innovation is the Structure Reference feature, available within the Text to Image module of the web application. This feature empowers users to generate images that not only adhere to the visual style of a reference image but also mirror its composition.
The inclusion of a new style engine within Image 3 allows for even more nuanced customization. This empowers users to subtly alter the visual direction of images generated from a base text prompt.
System Compatibility and Access Considerations
Image 3 is accessible through the web-based version of Firefly. This application offers compatibility with various operating systems and browsers. Specifically, it supports Chrome, Edge, Firefox, and Safari on Windows 10 (and subsequent versions), macOS 12.0 (and subsequent versions), and ChromeOS.
In terms of access, a credit-based system is employed. Notably, ‘fast’ generative credits can be obtained through subscriptions to Adobe’s Creative Cloud suite or via independent subscriptions specific to Firefly.