ID Photo Creation Tool

The online passport photo maker turns a portrait into a document-style photo with the required crop, background color, and size. It uses automatic background removal to isolate the person and supports common needs such as passport, ID, visa, application, school, and work badge photos. This is convenient when a formal image is needed quickly for online forms or internal records. Requirements can still vary by country, agency, and document type: head size, eye line, shadows, glasses, clothing, expression, and photo age may all matter. The tool helps prepare and preview a compliant-looking image, but official submission rules should be checked before relying on the final file.

Document-photo specs are dimensioned in millimetres at print and in pixels at upload, joined by the print DPI: pixel = inch × dpi, so a 25 by 35 mm one-inch print at 300 DPI is exactly 295 by 413 pixels, a 35 by 49 mm two-inch print is 413 by 579 pixels, and a US passport or visa 2 by 2 inch print is 600 by 600 pixels. Each preset stores the target width, height, and DPI tuple and the canvas is resized to those exact pixel dimensions before any further work. Resizing is done by drawImage on an OffscreenCanvas or HTMLCanvasElement, which on Chromium and Firefox uses a bilinear or trilinear filter for downscaling. Background replacement runs an in-browser segmentation model that outputs a soft alpha mask the same size as the source, so a uniform fill colour (white #FFFFFF, China visa blue #198CFF, or a custom hex) can be composited under the subject without leaving a halo. The final image is exported through canvas.toBlob('image/jpeg', 0.92) for paper-style backgrounds or 'image/png' when an alpha channel must survive. The failure modes are predictable. The exported file strips EXIF metadata, including orientation, so a portrait shot in landscape mode needs an explicit rotate before crop. The segmentation model is shipped as a WebAssembly bundle that has to download once (50-100 MB) and then runs from cache. And because every embassy publishes its own head-height window (US passport demands 25-35 mm head from chin to crown), pixel-perfect output is still only a first draft until checked against the official spec.

• Pixel-DPI bridge: width_px = width_in × dpi, so 1-inch (25×35 mm) at 300 DPI is 295×413 px and 2-inch (35×49 mm) is 413×579 px.
• Common presets: US passport/visa 2″×2″ at 600×600 px, EU/UK 35×45 mm, HK/Macau permit 33×48 mm blue (#198CFF), Schengen 35×45 mm white.
• Resampling: HTMLCanvas.drawImage uses bilinear filtering on downscale, so heavy crops should be done in a single drawImage call to avoid compounding interpolation loss.
• Background replacement: an in-browser segmentation model emits a soft alpha mask; the chosen fill colour is composited under the subject before export, no server round trip required.
• Export: canvas.toBlob('image/jpeg', 0.92) for paper-style submissions and 'image/png' when alpha must be preserved; both strip EXIF orientation, so rotate before crop.
• Spec drift: head-height windows vary by agency (US passport 25-35 mm from chin to crown, Schengen 32-36 mm), so the preset is a starting frame and the embassy rules are the binding check.