On Thursday, Adobe introduced the latest version of its Firefly series of image generation AI models, including a new model devoted to creating vectors and a redesigned web application that incorporates all its AI models along with some from competitors. Additionally, a mobile application for Firefly is in development.
Adobe’s newly launched Firefly Image Model 4 reportedly offers improvements over previous versions in terms of quality, speed, and control over aspects such as the structure and style of outputs, camera angles, and zoom capabilities. This model can generate images with resolutions up to 2K. Moreover, a more advanced version called Image Model 4 Ultra is available, capable of rendering detailed complex scenes.
Alexandru Costin, Adobe’s Vice President of Generative AI, stated that the models were trained using a higher magnitude of compute power, which allows them to produce more intricate images. Compared to the earlier generation models, these new versions enhance text generation within images and include features that enable users to create pictures in a specific style using their chosen images.
Additionally, Adobe has made its Firefly video model accessible to everyone, following its limited beta release last year. This model allows users to produce video clips from text prompts or images, specify camera angles, control shots with designated start and end frames, add atmospheric elements, and personalize motion design aspects. The generated video clips can reach resolutions up to 1080p.
The Firefly Vector Model has also been introduced, offering capabilities to produce editable vector-based artworks, and generate different variations of logos, product packaging, icons, scenes, and patterns.
Firefly’s web application now provides access to all these models as well as image and video generation models from other companies such as OpenAI (GPT image generation), Google (Imagen 3 and Veo 2), and Flux (Flux 1.1 Pro). Users have the flexibility to switch between models, and all images produced will have content credentials attached. Adobe indicated that additional AI models might be integrated into the web app in the future.
Furthermore, Adobe is publicly testing a new product called Firefly Boards, designed as a canvas for ideation or moodboarding. This feature allows users to generate or import images, remix them, and collaborate with others, akin to other AI-based ideaboards from companies like Visual Electric, Cove, or Kosmik. Firefly Boards are accessible via the Firefly web application.
The company announced that these new models will soon be integrated into its product lineup, although no specific timeline was provided for their rollout.
Adobe is also making its Text-to-Image API and Avatar API widely available and has announced the introduction of a new Text-to-Video API, currently in beta. These APIs are part of Adobe’s Firefly Services portfolio, which includes various APIs, tools, and services.
Additionally, Adobe is testing a web application named Adobe Content Authenticity to enable users to attach credentials to their creations for indicating ownership and attribution via metadata. Users can also specify whether AI companies are permitted to use their images for AI model training.
