Adobe has unveiled a significant update to its Firefly suite of generative AI tools, including the release of new image generation models, a redesigned web app, and upcoming mobile access.

Firefly image model 4

• The centerpiece of this update is the new Firefly Image Model 4, which Adobe says delivers improved image quality, faster generation speeds, and greater user control over outputs.

• This includes adjustments to composition, camera angles, zoom, and stylistic elements.

• The model now supports image generation at resolutions up to 2K.

Model 4 ultra

• A more advanced version, Image Model 4 Ultra, is also being introduced.

• This model offers enhanced rendering capabilities for complex scenes featuring fine details and small structures.

• According to Adobe’s Vice President of Generative AI, Alexandru Costin, these improvements were made possible by training the models with significantly more computing power.

• He noted that the latest models also offer better performance in text rendering within images and allow users to guide outputs using reference images for stylistic consistency.

Firefly video and vector models

• Adobe is expanding access to its Firefly Video Model, initially launched in limited beta last year. Now available to all users, the model enables text-to-video generation with support for custom camera angles, start and end frame selection, atmospheric elements, and motion design customization. Generated videos can reach resolutions of up to 1080p.

• The newly released Firefly Vector Model can produce editable vector artwork suitable for logos, product packaging, icons, patterns, and entire scenes. Users can iterate and create variations quickly, supporting professional-grade vector design tasks.

Redesigned firefly web app

• Adobe’s redesigned Firefly web app now consolidates all of its generative AI models in a single interface.

• The platform also features select third-party AI models, including OpenAI’s image tools, Google’s Imagen 3 and Veo 2, and Flux’s Flux 1.1 Pro.

• Users can easily switch between different models and all generated content comes with attached content credentials for authenticity and traceability.

• Adobe is also testing Firefly Boards, a collaborative ideation and moodboarding tool integrated into the web app.

• Boards allow users to generate, import, remix, and share images, with functionality akin to platforms like Kosmik, Visual Electric, or Cove.

API and services expansion

• Adobe announced that its Text-to-Image API and Avatar API are now generally available.

• Additionally, the Text-to-Video API is being offered in beta. These APIs are part of Adobe’s

• Firefly Services, a suite of APIs, tools, and services aimed at developers and enterprise users.

• Adobe stated that these new models and tools will be gradually integrated into its broader Creative Cloud ecosystem, although no specific rollout timeline has been provided.