Models and Training

Adobe Firefly incorporates a variety of models, each aimed at addressing specific user requirements and applications. As a constantly evolving platform, Firefly's models undergo ongoing refinement and expansion to enhance their capabilities.

First Model

The first model, launched during the initial Beta phase, was carefully trained on a vast dataset that includes Adobe Stock images, open-license materials and public domain content with expired copyright. This model primarily focuses on generating image and text effects, while ensuring that the produced content is suitable for commercial use. The extensive library of Adobe Stock, which contains hundreds of millions of high-quality, professionally licensed images, serves as a key component in the training process, enabling Adobe Firefly to avoid generating content that infringes on intellectual property rights.

Future Models

As Adobe Firefly continues to develop, new models will be introduced and integrated into the platform. These upcoming models will utilize a combination of resources, technology, and training data from both Adobe and external sources. By incorporating diverse inputs and leveraging advanced machine learning techniques, Adobe Firefly aims to continuously improve its capabilities and deliver exceptional results for users.