Black Forest Labs releases FLUX Erase, its image element removal tool.

Black Forest Labs releases FLUX Erase, an AI tool that removes unwanted image elements and automatically reconstructs backgrounds via a single mask.

Black Forest Labs adds a tool called Erase to its FLUX editing suite, dedicated to removing unwanted objects, text, or details from an image. Its unique feature is that the model was specifically trained to treat erasure and reconstruction as a single task, whereas classic inpainting involves multiple steps. The background is automatically recomposed, including shadows, reflections, and missing areas, without leaving artifacts or halos.

Its operation relies on just two inputs: an image and a mask where the white area indicates what should disappear, with the rest of the image remaining intact. A free demo allows users to upload an image, paint the part to be removed with a brush, then initiate the removal. For developers, an API exposes an asynchronous endpoint operating on the same black and white mask principle, without the need for a text prompt since the erasure instruction is integrated into the model. A dilation parameter, adjustable from 0 to 25 pixels, adjusts the contour coverage.

The use cases demonstrated by the developer include removing an object placed on a plain background, deleting text superimposed on a photo, and erasing shadows associated with a masked element. The tool is a continuation of existing FLUX modules like Fill, Depth, or Canny.