On Thursday, Adobe unveiled the latest version of its Firefly suite, which includes advancements in AI image generation models, the introduction of a vector model, and a redesigned web application that now hosts both Adobe’s and its competitors’ AI offerings. A mobile app is also in development.
Firefly Image Model 4 boasts improvements in image quality, speed, and user control, enabling outputs in resolutions up to 2K. Additionally, the upgraded Image Model 4 Ultra can handle more complex scenes with intricate details. According to Alexandru Costin, Adobe’s VP of Generative AI, these models have been trained with significantly more computational power, resulting in enhanced text generation capabilities within images, alongside features that allow users to influence stylistic outcomes using their own reference images.
Adobe has also expanded access to its Firefly video model, previously in limited beta, enabling users to create video clips directly from text prompts or images. This model allows for customisation of camera angles, start and end frames, atmospheric elements, and motion design, producing videos in resolutions up to 1080p.
A new Firefly Vector Model offers the ability to create and modify vector art, facilitating the design of logos, product packaging, icons, scenes, and patterns with ease. The comprehensive Firefly web app brings together all of these models and includes additional image and video generation capabilities from notable competitors like OpenAI and Google, allowing users to switch between different AI models seamlessly. Images produced through the app will carry content credentials for ownership verification.
Additionally, Adobe is beta testing Firefly Boards, a collaborative platform for ideation and moodboarding, where users can generate or import images and work together creatively, emulating functionalities seen in other AI-based brainstorming tools.
These new tools are set to be integrated into Adobe’s wider product ecosystem, although the company has not specified when this will occur. Furthermore, Adobe is making its Text-to-Image and Avatar APIs available to the public and has introduced a beta version of its Text-to-Video API as part of its Firefly Services suite.
In a move towards ensuring content ownership, Adobe is also testing a web application called Adobe Content Authenticity, which enables users to attach credentials to their work, detailing ownership and enabling the setting of permissions regarding the use of their images for AI model training.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
