Adobe Firefly Revolutionizes Creation with 30 Integrated AI Models

Adobe Firefly Integrates Over 30 AI Models

Adobe recently unveiled a major update to its Firefly tool, which now features over 30 artificial intelligence models accessible from a single platform. Among these models are prestigious names such as Veo 3.1 and Nano Banana 2 from Google, Gen-4.5 from Runway, as well as generative models from OpenAI and Kling 2.5 Turbo. This integration allows users to switch from one model to another without leaving the interface, providing increased flexibility and efficiency.

At the same time, Adobe announced that its own model, Firefly Image Model 5, is now available to everyone. This advancement enables creators to generate images with OpenAI, refine them with Runway, and then edit them using Adobe's tools, all without leaving the application. This unique approach positions Adobe as a key player in the realm of professional creative tools, offering unlimited generations of images and videos through this selection of partner models. This significantly alters the cost dynamics for those producing at scale, as they can now access these resources without restrictions.

Adobe stands out by aggregating models from other companies like Midjourney, Runway, and OpenAI, which is a unique position in the market for professional creative tools. This strategy allows Adobe to offer a service that no one else currently occupies, further solidifying its leadership position in the sector.

Train a Model on Your Own Style

Adobe has also introduced custom Firefly models, currently in public beta, which allow users to train the AI on their own visuals. This means that creators can replicate their unique aesthetic, whether it involves illustration styles, character consistency, or specific photographic renderings. By importing their assets, Firefly analyzes and generates a model that aligns with this visual universe.

Three main use cases are highlighted by Adobe: illustration styles where precision of lines and color consistency are essential, recurring characters that must remain identical across different scenes, and photographic renderings to be applied to large quantities of images. This feature is particularly beneficial for studios or agencies managing multiple campaigns, as it allows them to reuse a trained model from one project to another without starting from scratch.

Regarding data ownership, Adobe assures that these models are private by default and that the generated content fully belongs to the creator. This clarification is crucial in a context where the legality of data usage remains unclear for many competing tools, such as Midjourney and Runway, which do not offer such explicit guarantees.

Project Moonlight and Quick Cut

Project Moonlight, an agentic assistant unveiled at the Adobe MAX event in October 2025, is now in an expanded private beta. This tool works with Photoshop, Express, and Acrobat, allowing users to give natural language instructions to perform actions within the application. Creators can then refine and enrich these actions, although Adobe has not yet announced a date for general availability.

Additionally, Quick Cut, a new feature, transforms raw footage into a structured first edit in just a few minutes. Combined with image editing capabilities, such as adding or removing objects, extending scenes, and fine-tuning adjustments, Firefly now offers an almost complete production pipeline, from initial generation to final rendering ready for distribution.

Two years after its launch in 2024, Firefly has evolved from a simple image generator to a multimodal production environment, where competing models coexist harmoniously, thereby redefining the standards of digital creation.