The short answer
Use the background remover in two steps: cut out to transparency, then fill with the target solid colour. The buttons are trivial; the craft sits in two parameters. <strong>Tolerance</strong> decides how close a pixel must be to the background colour to count as background; <strong>feathering</strong> decides how soft the edge transition is. The defaults work for most white-background ID photos, but when hair looks grey or the background survives, one goes up and the other goes down — section four has the recipe. And pick the background colour from the submitting page's wording, not from what other people happened to upload.
Which background for which occasion
White is the default: passports, most civil-service and teacher-certification exams, CVs, diplomas, social-security cards, and nearly every staff badge. Blue shows up on some professional exams (construction, safety certification), work badges, some older diplomas, and certain licence or onboarding documents. Red is the rarest — mainly some title-review applications, party-membership paperwork, or specific commercial cases. Three working rules: the wording on the submission page is the only authority; different systems in the same packet may ask for different colours, so keeping two variants is normal; and some systems check the actual pixel colour rather than eyeballing it, so use a genuine solid white instead of "almost white" light grey.
A bad original cannot be rescued
Swapping a background fixes the background, not the photo. Reshoot or retouch first if the original has: a non-solid background (gradient, shadow, clutter, door frames, other people's clothing); a subject whose colour is too close to the background (off-white wall with a white shirt, white wall with a pale top); uneven lighting; a profile shot; glare on the face; or heavy noise. The reason is mechanical: cutting out relies on colour difference, and where that difference is thin, no parameter setting saves you from eating the hair or the shirt. Most algorithms only handle solid or near-solid backgrounds anyway — a cluttered scene gets swallowed whole.
Tuning tolerance and feathering
Raising tolerance widens the colour band treated as background: cleaner cutout, but parts of the subject that resemble the background get eaten too — the classic symptom is transparent hair tips or a missing patch of clothing. Lowering it leaves background residue: a grey rim or a white outline around the subject. Raising feathering softens the edge naturally but makes hair look blurry; lowering it keeps edges crisp but shows jaggies when zoomed. Tune in this order: <strong>first tolerance</strong> until the background is gone and the subject is intact, <strong>then feathering</strong> until the edge looks natural. A common working point is tolerance 30–40% with 1–3 px of feathering. Judge by zooming the result to 200% and looking at the hairline and shoulder line — no white rim, no bites out of the subject, individual hairs still visible.
Four checks after swapping
One, edges: the hairline, the ear outline, and the shoulder-to-clothing boundary — these three give the game away. Two, fill: look for patches of the old background showing through the new colour, especially where hair is thin and around shoulder corners. Three, colour values: sample the corner pixels with an eyedropper rather than trusting your eyes. Four, size and weight: after exporting, confirm the file size is still inside the limit — a background change does not shrink the file, and PNG photos with a solid backdrop are often much larger than JPEG, so export JPG rather than PNG when the cap is 200 KB.
The full chain
The pipeline runs: crop to spec first (one inch 295×413 px, two inch 413×579 px) → remove the background and apply the new colour → resize to the required dimensions → compress to the target weight → export. The order matters: cropping before cutting out gives clean edges, and resizing after cutting out avoids introducing fresh jaggies into already-processed edges. Tools: <a href="remove-bg.html">background removal</a> handles transparency and direct white/blue/red output, <a href="image-crop.html">cropping</a> and <a href="image-resize.html">resizing</a> handle geometry, and <a href="image-compressor.html">compression</a> handles the weight. All of it runs locally, so ID photos never touch a server.
想马上动手：用 在线抠图 换底色，尺寸不对去 图片裁剪，体积超标用 图片压缩 压到目标大小；还没拍证件照的先看 证件照怎么拍。Try it now: swap the background with background removal, fix the crop via image cropping, and hit the size limit with image compression; if you have not taken the photo yet, start with how to shoot an ID photo.
FAQ
After swapping, the hair edge looks white or grey.
Classic symptom of too much feathering, or residue left by the cutout. Reduce feathering (1–2 px) to tighten the edge; if a white rim survives, the background was not fully removed, so raise the tolerance a little and retry. Remember the division of labour: tolerance handles cleanliness, feathering handles softness. Tune them one at a time instead of moving both at once.
My background is a gradient or has shadows — can it still work?
Sometimes: gentle gradients often come through, while distinct shadows and colour patches leave residue or bite into the subject. The reliable route for a gradient background is to reshoot — find a plain surface (white wall, white paper, white sheet) and avoid casting a shadow. With even light, one pass does it. Forcing a hard cut on a bad background takes longer and looks worse than reshooting.
The system rejected my photo after the background swap.
Most systems check dimensions, format, file weight and the pixel colour of the background — not whether the colour was swapped later. So a rejection almost always comes down to one of four things: the colour is not the exact required value (sample the corners), the dimensions or aspect ratio are off, the file exceeds the limit, or the format is wrong (JPG and PNG rules differ). Check the submission notes line by line; usually fixing one item passes it.