Erase solid-color image backgrounds locally in your browser and export transparent PNG, with tolerance and edge feathering controls
Click to choose an image, or drag and drop it here
Supports JPG / PNG / WebP / GIF / BMP, or paste with Ctrl+V. Everything is processed locally and never uploaded.
Samples the image corners to work out the background color automatically, removing it in a single click
Click anywhere on the image to add that color to the erase list; stack several colors for gradient backdrops
Tolerance controls how aggressively color is removed; feathering and shrink keep the outline natural
Export a transparent background, or switch to white, blue or red in one step for ID photo backdrops
Your image stays in the browser and is processed with Canvas; nothing is uploaded and no account is needed
Save as PNG (transparent), JPG or WebP with adjustable quality, or copy the result straight to the clipboard
Click the upload area to choose a file, drag an image onto the dashed box, or paste a screenshot with Ctrl+V. The file is only read locally.
Click Auto Remove: the tool samples the four corners to determine the background color and removes it in one pass. The preview updates instantly.
If background remains, use the eyedropper on those areas or raise the tolerance. If the subject edge is being eaten, lower tolerance and increase edge feathering.
For ID photo backdrops, choose white, blue or red under Background Fill, then pick the export format and click Download Image.
Switch a blue or red ID photo to a white or custom backdrop for applications and official documents
Remove the shooting background and use pure white to meet marketplace main-image requirements
Drop the white box behind a logo, stamp or signature so it overlays cleanly on posters and contracts
Erase grey or noisy backgrounds from scans and slide screenshots so the content stands out
Cut the subject into a transparent PNG and reuse it as a layer in your design tools
Keep ID cards and contracts off third-party servers by processing them entirely on your own machine
Digital images are made of pixels, each carrying red, green and blue values plus an alpha (transparency) channel. Removing a background means deciding which pixels belong to the background and setting their alpha to 0. For solid backdrops the approach is straightforward: estimate the background color, then measure the color distance from every pixel to that color and treat anything beyond a threshold as background.
Tolerance is that color-distance threshold. Too low, and parts of the background that differ slightly from the sampled color will remain; too high, and pixels of the subject that resemble the backdrop get erased. Start around 20-40 for clean solid colors; raise it to 50-80 for slight lighting gradients. If the subject itself is close to the backdrop (a white shirt on white), keep the tolerance low and patch the leftovers with the eyedropper.
A hard threshold leaves jagged edges and a rim of background color. Feathering makes the transparency transition gradually near the boundary so the outline looks soft and natural, while shrink eats a little inward to remove that leftover colored rim. Used together they noticeably improve cutout quality on solid backgrounds.
A color-based algorithm can only reason about color. Once the background contains many colors (scenery, interiors, gradient light spots), or the subject is close in color to the backdrop, or there is fine semi-transparent hair at the edges, color rules cannot reliably separate foreground from background. Those cases need a semantic-segmentation AI model rather than a purely client-side color algorithm.
| Name | Hex | Typical use |
|---|---|---|
| Pure white | #FFFFFF | ID photo white backdrop, marketplace main images |
| Standard blue | #438EDB | Common ID photo blue backdrop |
| Deep red | #D9001B | Common ID photo red backdrop |
| Sky blue | #00B4FF | Graduation and staff photos |
| Light grey | #F2F2F2 | Neutral backdrop for product and model shots |
| Transparent | Alpha=0 | Assets, logos and images meant for compositing |