Tutorial · 9 min read · Updated 14 August 2026
How to remove photo backgrounds cleanly for e-commerce and IDs
Get crisp, transparent cutouts around hair, fur, and intricate edges using on-device machine learning without sending photos to a cloud server.
A clean product cutout or a passport photo that meets the rules both need the subject separated from the background. This used to mean slow, manual work with the pen tool in expensive graphics software. Now machine learning models running in the browser can separate even hard subjects, like loose strands of hair and fuzzy edges, in a few seconds. That is what Remove Background runs, on your own device.
Key Challenges in Automated Background Removal
An automatic background remover has to keep the subject's edges sharp and still handle soft, partly transparent areas:
- Edge Feathering vs. Alpha Matte: A hard cutout around hair or fabric leaves an ugly, jagged halo. Good background removal adds an alpha matting pass, which blends transparency at the sub-pixel level so the edge looks natural.
- Similar Contrast Boundaries: When someone in white clothes stands against a light grey wall, colour alone cannot separate them. The model has to rely on what it has learned about the shape of the human body.
- Color Spill Removal: Strong background colours (a green screen, a bright blue wall) reflect coloured light onto the edges of the subject. Good tools reduce the saturation of those edge pixels to remove the coloured halo.
Preparing Photos for Best Background Removal Results
- Make Sure the Photo Is Sharp: Blurry edges, or edges smeared by movement, make it hard for the model to find where the subject ends.
- Keep the Subject Away from the Background: Place the subject at least 3 feet from the background. This cuts down shadows and reflected light.
- Export as 32-bit Transparent PNG: Save the cutout as a PNG with a full 8-bit alpha channel, so you can place it on any background, brochure or online shop page. If the cutout then needs to fit an exact frame, such as a passport photo or a marketplace listing, finish with Crop Image.
On-Device AI: Your Portraits Stay Private
Employee badge photos, ID cards and unreleased product mockups are often confidential. The background remover runs small ONNX models inside your browser, using ONNX Runtime Web with WebAssembly and WebGL. Your photos never leave your device.