Sample one colour from a photo, choose how close other pixels must be, and paint them with a new shade. Tolerance and soft-edge controls decide how far the replacement reaches and how gently it fades into the untouched areas, while a live readout tells you what share of the image was changed.
This is a colour-key tool for flat backgrounds, product backdrops and simple graphics rather than a full recolouring studio. All processing runs locally in your browser. Your data is not uploaded to any server. 100% Local
Tolerance is a radius in RGB space: a pixel joins the replacement only if its distance from the sampled colour falls inside that radius. Soft edge blending then turns the outer slice of the radius into a gradient, so pixels just outside the core are mixed instead of switched. A hard edge with zero softening is right for logos and flat graphics, while a 20â40% blend hides the stair-stepped fringe that appears on photographs.
Matching the brightness of the original keeps the image believable; a much darker replacement flattens shadows and looks pasted on. Pull the replacement slightly towards the dominant light colour in the scene, and avoid pure saturated hues on large areas. For a white product backdrop, #F7F7F5 often reads better than #FFFFFF because it keeps a hint of shading instead of burning out.
Catalogue rules often demand a clean white backdrop. Sample the grey studio wall behind the product, raise tolerance until the wall is uniform, add 20% edge softening to remove the fringe, and set the target to #FFFFFF. Products with reflective edges may still keep a thin grey line, which is a sign to lower tolerance rather than push it further.
A single-colour garment, chair or phone case can be recoloured without touching the rest of the frame. Pick a mid-tone area rather than a shadowed fold, keep tolerance near 20%, and keep the replacement close in brightness. Folds and highlights stay darker or lighter, which is what makes the new colour look photographed rather than painted.
Brand assets exported as PNG often need a one-colour variant for dark or light backgrounds. Set tolerance to 0â5%, softening to 0, and replace the exact brand colour. Because the fill is flat, every pixel matches at once and the result stays perfectly crisp with no halo, which makes this the most reliable use of the tool.
Scans and screenshots often carry an off-white or pale blue background that looks dirty in a document. Sample the background once, use a wide tolerance of 30â50% and a soft edge of 30%, then replace it with white. Text and lines keep their edges because they sit far outside the match radius, but faint anti-aliasing may need a second pass.
The halo is made of pixels that blend the object and the background, such as anti-aliased edges or JPEG noise. They sit outside the matched range, so they keep the old shade. Raise tolerance slightly and add soft-edge blending, or replace a colour that is further away from the object.
Tolerance is the radius around your picked colour, measured in RGB space and scaled so 0 percent matches only an exact value while 100 percent matches every pixel. Half of the range already covers most variations of a flat background, so start near 20 percent and increase only if patches remain.
It works, but JPEG compression scatters neighbouring shades across a range, so a flat wall may contain dozens of slightly different colours. A tolerance of 25 to 40 percent catches most of them. Expect some speckle on very noisy files; a PNG or a larger original behaves more predictably.
One source colour is handled per pass. To change two colours, replace the first, download the result, then reopen that file and pick the second colour. Keep the first replacement in mind because the new file is the new starting point.
No. The pixels are read into a canvas in your own browser and the replaced image is encoded there too. Nothing is sent to a server, which also means the tool keeps working on a plane or on a locked-down office network.