Cut a subject out of a plain background by colour, feather the edge, and keep it transparent, fill it with a colour, or blur what was behind.
Reads the colours along the four edges of the picture, groups them into up to three background colours (or uses one you pick on the photo), and makes the matching area connected to the edges transparent, so a white shirt on a white wall stays whole. Switch off Only from the edges to remove every matching pixel. It is a colour-key cut-out, not a machine-learning one: it works well when the subject sits on a plain, evenly lit backdrop such as a product shot, a logo on white, a sticker on a green screen or a document photo, and it cannot separate a person from a busy street. The edge of the cut can be feathered, and the removed area can stay transparent, be filled with a flat colour, or show a blurred copy of the original scene.
A 2000 × 2000 photo of a coffee mug on a white tabletop: with Sensitivity 35 and Colour range 50 the white goes away in one pass, and Edge softness 2 hides the halo around the handle. Saved as PNG the cut-out is about 1.1 MB; as WebP at quality 90 it is around 240 KB and still transparent. A black logo scanned from paper needs Sensitivity 20 and softness 0 so the thin strokes stay crisp. A subject photographed against a green sheet with folds needs Colour range 80 so the darker creases still count as background.
Use it for product photos, logos, stickers, screenshots of a chart on a flat colour, or anything on a green screen. Skip it for portraits against a room, landscapes and group photos; the background there is not one colour and the result will have holes. If the subject touches the edge of the frame, the tool may read the subject as background, so crop a little air around it first with Crop Image. After the cut, Round Corner or Clip Image can shape the transparent canvas, and Add Text can drop the result onto a caption.
Sensitivity is how far a pixel may be from a background colour and still be removed; raise it when specks of background remain, lower it when parts of the subject vanish. Colour range controls how the edge samples are grouped into background colours; wider handles gradients and uneven light, narrower protects a subject that shares a tone with the backdrop. Edge softness feathers the cut in pixels: 0 for logos and pixel art, 1 to 2 for photos, up to 5 for hair-like edges. Background Colour fills the removed area, useful for putting a product on a brand colour, and Blur keeps the scene but softens it with the radius you set. Transparent output needs PNG, WebP or AVIF; JPEG fills transparency with the Background colour in the Output section.
Why did part of my subject disappear? It shares a colour with the backdrop. Lower Sensitivity first, then narrow the Colour range. If a white shirt sits on a white wall, no colour-based cut can tell them apart; photograph it on a contrasting sheet.
There is a thin outline left around the edge. How do I remove it? Raise Sensitivity by 5 to 10 and set Edge softness to 1 or 2. The outline is the anti-aliased mix of subject and background colour along the boundary.
Can it remove the background from a portrait? Only when the person is in front of a plain wall or a green screen. Real rooms and outdoor scenes contain too many colours for a colour-key cut.
Does the photo get uploaded? No. The cut-out is computed on your device. You can turn off the network after the page loads and it still works.