Remove Background from Image: Free, Private, No Upload

Drop image or click to upload

PNG, JPG, WebP, GIF, BMP

Files never leave your device

How do I remove a background from an image?

Drop your image and click Remove Background. The AI model runs entirely in your browser, so your photos stay on your device. Processing takes a few seconds on most desktops. The model downloads once and is cached by your browser for future use.

How on-device background removal works

This tool uses open-source segmentation models running through Transformers.js. On browsers with WebGPU support (Chrome, Edge), it uses BiRefNet for the best quality. Otherwise it falls back to MODNet, which works everywhere. When you process an image, JavaScript sends pixel data to a Web Worker where the model identifies the subject and creates a transparency mask. No data leaves your device at any point.

What images work best

Portrait photos: Clear subject separation from the background works well for headshots, team photos, and profile pictures.

Product photos: Items on solid or simple backgrounds produce clean cutouts for e-commerce and listings.

Pet photos: The model handles animal outlines, though fine fur edges may need the softness slider.

Graphics and logos: High-contrast graphics with distinct foreground elements.

Edge softness control

After removing the background, use the edge softness slider to feather the transition between subject and transparency. A value of 0 gives hard edges (good for graphics). Higher values create softer, more natural edges (better for photos that will be composited onto new backgrounds).

Free vs Pro resolution

Free users get background removal with output capped at 1024px on the longest side. That covers profile pictures, social media posts, and web graphics. Pro users get full-resolution output up to 4096px, suitable for print and high-resolution displays.

Does this tool upload my images?

No. The AI model runs directly in your browser using Transformers.js, with WebGPU acceleration when available and WebAssembly as fallback. Your images never leave your device. You can verify this by watching the Network tab in DevTools while processing.

How big is the AI model download?

The model size depends on your browser. Chrome and Edge with WebGPU support download BiRefNet (~200MB) for the best quality. Other browsers (including Firefox) download MODNet (~24MB), a lighter model that runs on CPU. Either way, the model downloads once and your browser caches it for future visits.

What browsers are supported?

Chrome, Firefox, and Edge all support the Web Worker and WebAssembly features needed for on-device AI. Browsers with WebGPU support can use faster GPU inference. Safari support is improving.

Can I use this on my phone?

Yes, but processing takes longer on mobile devices (10-20 seconds). Large images may also use significant memory. For best results on mobile, use images under 2000px.

All processing happens on your device using a Web Worker and the Canvas API. The AI model runs locally via WebGPU when available, otherwise WebAssembly. Your images are never uploaded, and no data is sent to any server.