.png)
April 28, 2025
Adobe has comprehensively revised and expanded Firefly, its AI platform. Originally designed as an image generator, the tool has now become a versatile system for creating diverse digital content, capable of generating images, videos, audio, and vector graphics. Since its introduction almost two years ago, over 22 billion pieces of content have been created worldwide with Firefly, a success Adobe aims to build upon with this update.
With the Firefly Image Model 4, Adobe introduces two new versions: Standard and Ultra. According to Adobe, the Standard model covers 90 percent of common creative needs, while the Ultra version is designed for more complex, photorealistic scenes. Both models are said to offer improved representation of people, animals, and architecture compared to previous versions. Text rendering on images has also been optimized, and the models are supposed to respond more accurately to prompts. The text-to-image function offers enhanced control options, including aesthetic filters, the selection of specific styles, and options for precise adjustment of compositions.
In addition, Adobe has officially released the Firefly Video Model, which can generate video clips up to five seconds long in various resolutions and aspect ratios. The company promotes this model as "commercially safe" because it was trained on licensed Adobe data. Users of the Firefly Premium subscription receive unlimited access to the Firefly Video Model in all Firefly apps. The system can generate video clips up to five seconds in length, with various resolutions up to 1080p and multiple aspect ratios, including 16:9, 9:16, and 1:1. Adobe emphasizes the improvements in photorealism, detail, and the representation of text, landscapes, and visual effects. Users can influence video creation by uploading start and end images or seamlessly switching between text, image, and video inputs. Camera movements can also be customized.
Firefly's new features include a "text-to-vector" module for creating editable vector graphics, "Firefly Boards" for collaborative concept development, and the already introduced "translate audio" function for language translations that retain the original voice. To improve mobile use, Adobe announced an app for iOS and Android devices. This mobile version will enable the creation of images and videos on the go and support project synchronization between mobile and desktop platforms.
Adobe is expanding its platform by integrating AI models from third-party providers. Models such as Google Imagen 3, Veo 2, OpenAI GPT Image Generation, and Flux 1.1 Pro are to be integrated into the Firefly ecosystem. Future integrations will include models from fal.ai, Runway, Pika, Luma, and Ideogram. To ensure transparency, Adobe will mark all AI-generated content with "Content Credentials" that indicate which model was used for creation. However, the company emphasizes that only the proprietary Firefly models are trained exclusively on licensed data to ensure commercial use. Some third-party model providers, such as OpenAI, are currently involved in legal disputes over their training data.
