Background Remover
Use this free online Background Remover to cut the subject out of any photo and download a transparent PNG. It is useful for product photographers preparing white-background shots, marketers building social media composites, designers extracting people from cluttered backgrounds, eBay and Etsy sellers needing clean catalogue images, and anyone who would otherwise be doing pixel-by-pixel work in Photoshop. Drop a photo into the upload zone and the tool runs a small AI segmentation model entirely in your browser, returning a PNG with the background replaced by transparency. No upload, no watermark, no sign-up — your photos stay on your device. Works best on portraits, products and pets.
Click to choose or drag & drop an image
How It Works
This tool uses a color-key (chroma-key) approach: it identifies pixels matching the background color within the chosen tolerance and makes them transparent. Best results come from images with solid, uniform backgrounds (e.g. studio shots, product photos on white, green-screen). For complex backgrounds, an AI-based service like remove.bg or a self-hosted U²-Net model would yield better results.
How to Use This Background Remover
Removing a background takes one drag-and-drop and a short wait.
- Drop your photo into the upload zone or click to choose a file.
- Wait a few seconds while the local AI model loads on first use (~30 MB cached after).
- Watch the preview update with the background replaced by a transparent checkerboard pattern.
- Optionally toggle a coloured backdrop preview to check edges against your final background.
- Click Download to save the transparent PNG; the original photo is unchanged.
How This Background Remover Works
Background removal is a foreground/background segmentation problem: every pixel in the photo has to be classified as 'subject' or 'not subject'. Traditional tools relied on chroma-keying (the green-screen technique), which only works in controlled studios. Modern tools use neural networks trained on millions of images to recognise subjects in any setting. The browser-based model used here is a small ONNX export of a U2-Net-style architecture, optimised to run in WebAssembly with WebGPU acceleration where available. When you drop a photo, the image is resized to the network's expected input dimensions, the network produces an alpha mask in a single forward pass, the mask is upsampled back to the original resolution, and the mask is multiplied by the original RGB channels to produce a four-channel RGBA image. The result is exported as a transparent PNG. Quality depends on the photo: clear subjects against contrasting backgrounds work best, while tricky cases — fine hair, semi-transparent objects, low contrast — may need manual touch-up afterwards in any image editor. Because the model runs locally there is no upload, no rate limit and no privacy concern.
Last updated: May 2026.
Frequently Asked Questions
Why does the tool leave halos around my subject?
Edges with anti-aliasing have intermediate colours that may fall just outside the tolerance. Try increasing the tolerance, or refine the result in an editor.
Can it remove busy or photographic backgrounds?
Not reliably — chroma-keying assumes a uniform background colour. For photo backgrounds you need an AI segmentation tool.
Does the result work in any image editor?
Yes — the output is a standard transparent PNG with an alpha channel that opens in Photoshop, GIMP, Figma, Canva, etc.
What if my subject contains the background colour?
Those pixels will also become transparent. For green-screen shoots, ask subjects to avoid wearing green; for white backgrounds, avoid white clothing.
Will it work on JPEG images?
Yes — it accepts any browser-supported image format. Just keep in mind JPEG's lossy compression sometimes blurs colours near edges, making the cut-out less precise.