Free · No sign-up · PrivateSmart / Local Tools
Detect the objects in an image, click the subject you want, refine the mask edge, then export either a black-and-white mask or a transparent PNG.
Tip: detection selects the area; segmentation separates the subject inside it. Thin edges, glass, fur and overlapping objects can still need a close check.
Private by design: both local models run in this browser. First use may download them.
A mask is a simple black-and-white image that tells another design or editing tool which pixels belong to the selected subject. It is useful for compositing, selective effects, generative-fill workflows, colour adjustments, motion graphics and any job where you want a clean subject selection without manually tracing from scratch.
The first local model finds likely object boxes. You choose the subject by clicking a box or selecting it from the list. A second local model then tries to separate the main subject inside that selected area. That two-step flow is much more useful when an image contains several objects than simply masking whichever subject the model thinks is most important.
Use Selection padding if the detector box is too tight. Expand / contract moves the mask outward or inward by a few pixels, which is helpful around product edges or when a cutout includes a faint background fringe. Edge softness feathers the transition. Invert flips which side is treated as the selected area.
Hair, fur, glass, smoke, thin wires, shadows and overlapping subjects can confuse automated segmentation. Always inspect those areas closely before using the mask for final production work.
Useful to know
Short, practical guidance for this specific kind of task, without turning a tiny job into a tutorial marathon.
What this tool does
Detect and select an object, refine the mask edge, then export a black-and-white mask or transparent PNG locally.
The tool is designed to get that one job done without making you learn a full piece of software first.
Best results
Inspect hair, fur, glass, thin objects, shadows and places where subjects overlap. Automatic segmentation can look convincing at fit-to-screen size while still missing small edge detail.
If the object is clipped, increase Selection padding. Use Expand / contract for small fringes and Edge softness only as much as the destination needs.
In plain English
The detector chooses the object area first, then a separate local model builds the subject mask inside that area. You stay in control of which object is selected and how the edge is refined.
AI tools download their model on first use and local-processing tools work directly in your browser. Choose the specialist tool that best matches your subject for cleaner results.
File size & formats
JPG, PNG and WebP are the safest choices. Other image formats work when your browser can decode them.
There is no extra OutMonkey upload-size cap in the browser interface. Very high-resolution images can still hit your browser or device memory limit, especially when enlarging them or running local models.
If an enormous image will not process, use a smaller copy or reduce its pixel dimensions first. Keep the original untouched.