bria/extract-object
Natural-language description of the object to act on, e.g. 'the red car'.
Source image as a public URL, a base64 data URI (data:image/png;base64,...) or raw base64-encoded image bytes.
Run the model to see the result here.