FLUX 3 Image Lets You Art Direct With Boxes, and That Is the Right Interface

Black Forest Labs has launched FLUX 3 Image, and the interesting part is not the model but the way you talk to it. You drag out a box for every element that matters, describe what goes in it, and add a one-line scene description. The model renders each element inside its box.

The canvas runs on a 0 to 1000 grid whatever the aspect ratio, and each box is specified as [y_min, x_min, y_max, x_max]. You can make several targeted edits at once and everything you did not touch stays where it was. It supports up to 10 reference images and generates natively at 2K and 4K.

This is the right interface for our line of work. Art direction is a conversation about placement: this goes top left, that sits behind the product, leave the rest alone. Describing position in prose has always been the weakest part of AI image tools, because the model treats "left" as a suggestion.

Boxes turn a vague wish into a layout, and layouts are what clients can actually approve. They also make revisions cheaper, since a change request becomes a moved rectangle rather than a rewritten paragraph.

The page does not give pricing, so check the dashboard or sales before you plan a project around it. The caution we would add is that a controllable layout does not make the picture good. Taste still decides where the boxes go.