FLUX 3 Image

The FLUX 3 Image model, developed by Black Forest Labs, excels at generating and editing images using a unique approach. Users input bounding boxes and a scene prompt, and the model takes care of the rest, rendering the image at full resolution with intricate details preserved.

To create an image using FLUX 3 Image, users place every element on a canvas by defining a bounding box for each item. They then describe the content of each box and provide a single line that ties all elements together in a global caption. The model generates an element table, a JSON array containing information about each element, including its id, bounding box coordinates, and a description. This table allows users to edit individual boxes as needed, with all untouched elements remaining unchanged.

The model's native understanding of image layout and composition enables users to create well-composed images with just a text prompt. It can handle various composition types, such as set pieces around photographs, collages, panel grids, editorial spreads, or crowded scenes where every face has its designated place. To create an image, users can either manually draw boxes in the Playground or send a layout prompt to FLUX 3 through the BFL API.

In summary, FLUX 3 Image is a powerful tool that simplifies the process of image generation and editing by allowing users to define boxes and semantic ids for each element in the scene. The model then renders the image according to these specifications, providing a high-quality result at full resolution.