The new model positions each element on a coordinate grid, edits a specific area without affecting the rest, and combines up to 10 reference images.
Black Forest Labs has presented FLUX 3 Image, the image generation and editing component of its FLUX 3 multimodal family. Its main new feature is an explicit layout: the user draws boxes on a canvas marked from 0 to 1,000 on each axis, regardless of the format, describes the content of each one, and then adds a sentence that connects the whole.
Localized retouching and multiple references
Each box can be modified separately: you can change an element without the neighboring areas moving one pixel. The model accepts up to 10 reference images, designated in the order they are added, and decides on their placement and size. It produces natively in 2K and 4K, up to 5,456 × 3,072 pixels, or 16.8 megapixels.
FLUX 3 Image can be accessed in a free playground and via BFL's API; commercial weights are licensed, and open-weight versions are released on Hugging Face and GitHub under a non-commercial license.
Read the full article on the original source
Go to the source →