bria/extract-object
Natural-language description of the object to act on, e.g. 'the red car'.
Source image as a public URL or raw base64-encoded image bytes (no data-URI prefix).
Run the model to see the result here.