Firefly: Adobe transforms her app into line into a creative war machine doped at AI

The mastodon of the Adobe creation software obviously has things to tell about the improvements made to Photoshop, Illustrator or Lightroom, its historic software.

But during his conference Adobe Max 2025 which took place on April 24, 2025 in London, Adobe opened the ball with the magic of the AI ​​and presented the many innovations integrated into its product which is most on the rise: Firefly. Not the AI ​​model in itself, which is currently in its version 4. But its online application, which becomes more and more powerful.

Third -party models arrive in Firefly

The most important of the announcements of the day is the arrival of third -party creation models in the Firefly online application. The most strong name is obviously that of Openai, which today sees its OpenGpt-4O model integrated into generation tools. So far, only Adobe models (Firefy Model 3, Firefly Model 4, etc.) were available. Safety tools (read below), but with more limited creative potential.

Famously made for the wave of images replicating the style of the Ghibli studio, the GPT-4O model is one of the most powerful and popular models. And today it is entering the interface of the Firefly online app (yes, Firefly is both the name of the app and the names of the AI ​​models).

However, the use of third -party models is not free. While the subscribers of the Creative Suite have 1000 credits of monthly generation with the Adobe models, the use of third -party models requires a subscription to Firefly.

The integration of third -party models is a controlled step in the opening of Adobe. On the one hand, we can only select only models approved by Adobe. Then Adobe has a major asset for its Firefly Model 3 and 4 models: legal protection that we have already told you about. If creatives and the general public can have fun enjoying power – not really legal – GPT -4O type models, many professional users should prefer to protect themselves from any suspicion of copyright violation using Adobe models.

If Adobe does not claim to be a major model supplier, the company claims to be the “one -stop shop” (One Stop Shop) to generate content.

Firefly Boards, an inspiration (almost) infinite board

Coming from the Project Concept presented during Adobe Max 2024, Firefly Boards is, in the words of Adobe a “Mood Board”. In French, a graphic or “inspiration board” room board. Accessible from the online interface of Firefly, this function makes it possible to create an unlimited number of plates whose contours are (almost) infinite.

Allowing the creation of universes or zones during creative reflection sessions, Boards incorporates many ergonomic functions that make life easier – “quality of life” functions, as adobe employees call them. Functions that seem simple, such as the automatic image arrangement to facilitate their viewing, but which “save valuable minutes”, as the company’s demonstrators pointed out.

An important element to note is that the integration of generative functions allows artists to carry out almost all their creative reflection operations in the Firefly interface, without necessarily using Desktop applications, the heart of Adobe’s historical activity. The most advanced users will have to be used for operations requiring more control and advanced functions. But as Adobe develops Firefly, the company seems to develop an online portal allowing the whole coming to create and organize the creative act without advanced knowledge.

For the time being, the board function is in beta and requires Registration on a waiting list.

Private models and mass production

In contrast to the very general public interface of a personal creation tool like Board, Firefly also incorporates a literally “industrial” dimension. A less known side of Adobe, but which represents an increasingly important part of its business. And which sometimes prefigures certain consumer features.

Certain functions, such as Firefy Production and Firefly API are resolutely turned towards businesses. The first makes it possible to simply and quickly generate hundreds of variations in design, facilitating the declination of an image for different platforms, languages ​​or target media. The second is the possibility of integrating Adobe tools into their tools – online cropping, detection of subjects, etc.

But the element that foreshadows a potential future function is Firefly Custom Model. A function that allows companies to create, from a set of internal visuals, a model of generative creation. Questioned by us, an Adobe executive confirmed to us that the company is studying the fact of offering the general public such a mechanism. Which would allow everyone – or at least less to artists with a volume of sufficiently supplied to generate a style. And thus create a personal IA generation model.

If the company “has nothing to officially announce on the subject”, the fact is that personal generative models could quickly become a reality. An evolution that could further accelerate the current “generative wave”.

Adobe does not forget the fundamentals (and puts AI everywhere)

The ever -increasing power of the IA tools forced Adobe to start its conference on these tools, which have both virtue of accelerating the workflow of users of its conventional tools, only to bring it new user profiles. Nevertheless, once the IA announces, Adobe has focused on its fundamentals: Photoshop of course, with creation tools mixing photo and video, but also Illustrator, with very important vector performance gains.

Within all its historical applications, Adobe has integrated a multitude of ergonomic improvements and access to the functions that the company describes “improvement in quality of life”. But more or less visible, the company also and above all continued the inexorable integration of IA tools. In Photoshop, we take advantage of a new action panel (in beta) which allows a novice to type what he wants to perform and let the application perform the action for him, without knowledge of the tools.

In Premiere Pro, it is the 4K definition that reaches the generative extension function. Or a research field that allows you to find a rush by typing the content you are looking for. Or the automatic generation of subtitles – and their translation in several dozen languages! As for Lightroom, a new neuronal filter is able to automatically create masks for landscape elements such as water, plants, sky, etc.

Whether in the cloud, in the form of models or services, or in its applications, AI is now on all floors of Adobe products. But the measured and thoughtful integration of these functions in its Desktop products shows that the company reflects each integration and prefers to infuse functions, rather than jostling. If he makes the headlines less than Openai and others, the American software giant could be the real winner of the IA revolution. Because if you can change a creation model with one click, it is much more difficult to change the habits of a workflow.