On Thursday, Adobe announced the launch of the latest iteration of its Firefly family of image generation AI models. This new release includes an advanced model for generating vectors and a redesigned web application that houses all its AI models, along with select offerings from competitors. Additionally, Adobe is developing a mobile app for Firefly, expanding accessibility for users.
The standout feature of this launch is the new Firefly Image Model 4. Adobe claims this model significantly enhances the quality, speed, and user control over image outputs. Users can now manipulate various aspects, including structure, style, camera angles, and zoom levels. The model is capable of generating images in resolutions of up to 2K, making it a powerful tool for creators.
Moreover, Adobe has introduced a more advanced version named Image Model 4 Ultra. This variant is specifically designed to handle complex scenes that comprise intricate details and numerous small structures, providing even greater creative possibilities.
According to Alexandru Costin, VP of Generative AI at Adobe, the company has employed a higher order of compute magnitude during the training process. This advancement allows the new models to generate more detailed images than their predecessors. Furthermore, the latest models show significant improvements in text generation within images and offer features that allow users to utilize their chosen images as a style reference for generating new pictures.
In addition to the image models, Adobe is also expanding access to its Firefly video model, which was previously available in limited beta. This model enables users to generate video clips from either text prompts or images. Users can specify camera angles, define start and end frames, and control various atmospheric elements to craft their shots. The model supports video generation at resolutions of up to 1080p, enhancing the potential for high-quality video content creation.
The newly introduced Firefly Vector Model allows users to create editable vector-based artwork. This model can iterate and generate variations for logos, product packaging, icons, scenes, and patterns, making it an invaluable resource for designers and marketers alike.
Adobe’s revamped web app provides users with access to all these new models, as well as several image and video generation models from notable competitors such as OpenAI (with its GPT image generation), Google (featuring Imagen 3 and Veo 2), and Flux (with Flux 1.1 Pro). This flexibility allows users to switch between models seamlessly, and importantly, all generated images will feature content credentials for transparency and authenticity.
Adobe has hinted that additional AI models may be integrated into the web app in the future. Furthermore, the company is publicly testing a new product called Firefly Boards, which serves as a collaborative canvas for ideation and moodboarding. Users can generate or import images, remix them, and collaborate with others, similar to functionalities found in other AI-based ideaboards from platforms like Visual Electric, Cove, or Kosmik. Firefly Boards are currently accessible through the Firefly web app.
In conjunction with these advancements, Adobe is making its Text-to-Image API and Avatar API generally available, with a new Text-to-Video API also launching in beta. These APIs are part of Adobe’s comprehensive Firefly Services collection, which includes various tools and services designed to enhance creative workflows.
Overall, Adobe's latest updates to the Firefly suite of AI models mark a significant step forward in image and video generation technology, offering users more control, improved quality, and a broader range of creative tools.
