Image Object Matting
Eliminate
Model Overview
Image Object Matting is a precision extraction model designed to isolate a specific object from an image while preserving fine, semantically-aware edge detail. Rather than separating a single foreground from its background, it identifies a target object based on its semantic meaning and extracts it cleanly, even from complex scenes containing many competing elements. Well suited for detailed compositing and selective editing workflows, it offers greater control in multi-object scenarios than standard background removal—while Background Removal is better suited to images with fewer objects where the entire foreground can be extracted in a single pass, Image Object Matting excels at pulling one chosen object out of a crowded frame.
Key Capabilities
Accurately extracts a specific object from its background and from other surrounding objects, with precise edge handling.
Supports a select box to designate the target object in scenes containing massive numbers of objects.
Recovers discontinuous regions of the same object based on semantic understanding, keeping separated parts of a single subject intact.
Model Pricing
Output MP
Credits
Getting Started
Endpoint
model—"Object"
Model-Specific Parameters
box— decimal, optional, default = [0, 0, 1, 1]The four numbers are the relative left/top/right/top edges, respectively (between 0 and 1). The more precise (tight) the box is, the better the expected result. When no box is passed, it will use [0, 0, 1, 1] as default (i.e. the entire image), and might misunderstand the target object.
For the highest-quality outputs, we recommend sending images with a single, main object in the foreground.
Examples
Last updated