Adobe launches an updated Firefly online application and new Firefly picture generation models.

On Thursday, Adobe unveiled the most recent version of its Firefly family of AI models for creating images, a vector generation model, and a revamped web application that contains all of its AI models in addition to several from rival companies. Additionally, a Firefly mobile app is being developed.

Adobe claims that the new Firefly Image Model 4 is faster, better, and offers more control over zoom, camera angles, output structure, and style than its predecessors. Up to 2K resolution photos can be produced by it. Image Model 4 Ultra is a modified, enhanced version of this model that can create intricate sceneries with numerous details and tiny buildings.

Adobe trained the models with a higher order of compute magnitude so they could produce more detailed images, according to Alexandru Costin, VP of Generative AI at Adobe. He went on to say that the new models contain capabilities that allow users to utilize photographs of their choosing to get the model to generate pictures in that style, and they also improve text creation in images when compared to models from prior generations.

Additionally, the company is opening up its Firefly video model, which debuted in restricted testing last year, to the public. Users can create video clips from an image or text prompt, adjust camera angles, manage shoots by specifying start and end frames, create ambient effects, and personalize motion design features. Up to 1080p video clips can be produced by the model from text.

The Firefly Vector Model, on the other hand, can produce editable vector-based artwork as well as distinctions in logos, product packaging, scenes, icons, and patterns.

Notably, all of these models are available through Firefly’s online application, along with a few models for creating images and videos from Flux (Flux 1.1 Pro), Google (Imagen 3 and Veo 2), and OpenAI (GPT image production). Images created from any of these models will have content credentials associated with them, and users are free to move between them at any time. The business made a suggestion that the online app might eventually include additional AI models.

Additionally, Adobe is making a new tool called Firefly Boards, a canvas for ideation or moodboarding, available to the public. Like other AI-based ideaboards from Visual Electric, Cove, or Kosmik, it allows users to create or import images, remix them, and work with others. Boards can be accessed via the Firefly online application.

IMAGE CREDIT:ADOBE

The company also announced that these new models would be added to its lineup of products shortly, although it did not specify when this would happen.

Along with making its Text-to-Image and Avatar APIs widely accessible, Adobe also announced that a new Text-to-Video API is now in beta testing. The company’s Firefly Services suite of tools, services, and APIs makes these APIs accessible.

Additionally, Adobe is piloting a web application called Adobe Content Authenticity that enables users to add credentials to their work in order to use metadata to show ownership and attribution. Users can also specify which photos AI businesses are allowed to use to train their models.