Hajra Naz
Adobe has unveiled the latest advancements in its Firefly family of AI tools, including powerful new image and vector generation models, along with a redesigned Firefly web app. A mobile version of the app is also in development.
Firefly Image Model 4: Higher Quality, More Control
The headline update is the launch of Firefly Image Model 4, which boasts significant improvements in image quality, generation speed, and user control. The model allows fine-tuning of structure, style, camera angles, and zoom, with image resolution support up to 2K.
A more advanced variant, Image Model 4 Ultra, further enhances detail rendering, especially for intricate scenes involving small elements and rich textures.
According to Alexandru Costin, Adobe’s VP of Generative AI, these upgrades are powered by higher compute resources, enabling better visual fidelity and improved text generation within images. Users can now also input a reference image to generate content in a similar style.
Firefly Video and Vector Models
Previously available in limited beta, the Firefly Video Model is now open to all users. It generates 1080p video clips from text or image prompts, allowing customization of camera angles, start and end frames, atmospheric effects, and motion elements.
Adobe also introduced the Firefly Vector Model, which can generate editable vector graphics. It’s ideal for creating and iterating on logos, packaging, icons, patterns, and entire scenes.
Unified Web Experience and Third-Party AI Access
The redesigned Firefly web app now serves as a hub for Adobe’s AI tools and also integrates models from other leading platforms, including:
-
OpenAI (image generation)
-
Google (Imagen 3 and Veo 2)
-
Flux (Flux 1.1 Pro)
Users can seamlessly switch between models, and all generated content includes content credentials to ensure transparency and authenticity. Adobe also indicated that more third-party AI models may be added in the future.
Firefly Boards, a shared canvas intended for ideation and moodboarding, is a new product that Adobe is now testing. This tool allows users to generate or import images, remix visuals, and work together in real-time, offering functionality similar to AI-powered ideaboards like Visual Electric, Cove, or Kosmik. Firefly Boards is accessible via the Firefly web app.
Adobe announced that its new AI models will soon be integrated into its broader product lineup, though no specific rollout timeline was provided.
Additionally, Adobe has made its Text-to-Image API and Avatar API generally available, while its new Text-to-Video API is now in beta. The company’s Firefly applications, a collection of tools, applications, and APIs built for creative processes, includes these APIs.
Adobe is also testing a new web app called Adobe Content Authenticity, which allows users to attach metadata credentials to their content for ownership and attribution. The tool also enables users to specify whether their images can be used for AI training purposes.
