The pixel size is a range, not a single number
A US passport photo is square, and its pixel dimensions have to land inside a band: 2×2 in (51×51 mm) · 600×600 to 1200×1200 pixels, square. Both ends are real. Below the floor the file is out of spec; above the ceiling it is out of spec too, and a photo correct in every other way fails on arithmetic alone.
Almost every page answering this question publishes the floor and calls it "the" size. That is how people arrive with a technically correct file and still get bounced: one number matched, and the shape, the weight or the head position missed. The whole specification, row by row:
| Photo size | 2×2 in (51×51 mm) · 600×600 to 1200×1200 pixels, square |
|---|---|
| File size | 54 KB to 10 MB |
| Format | JPEG |
| Head height | 50–69% of the photo's height (1 to 1⅜ in / 25–35 mm) |
| Eye height | 56–69% up from the bottom of the photo |
| Background | Plain white or off-white |
Two of those rows are not about the file at all. 50–69% of the photo's height (1 to 1⅜ in / 25–35 mm) and 56–69% up from the bottom of the photo are fractions of the frame, so they travel with the crop rather than the pixel count: scaling a compliant photo changes neither, and re-cropping one changes both.
The one sentence about pixels in the US manual runs the other way
8 FAM 402.1, the State Department's passport photo adjudication manual, is the document a US passport photo is adjudicated against. It contains exactly one sentence about pixels, and it is not a dimension:
| Pixels | "Photographs with visible pixels are not acceptable." |
|---|---|
| Grain | "Images composed of grain-like particles may not be acceptable." |
| Digital files | "Computer-generated photographs are acceptable if the image is sharp and clear." |
| Resolution | A stated goal of adjudication is that "the photograph is of sufficient resolution for facial recognition" |
| Overall size | "The photograph should be 2” x 2” (about 5 cm. x 5 cm.) in overall size." |
The manual does not count your pixels; it looks at whether the picture has fallen apart into them — soft blocks along the jaw, stair-stepping through the hair, a mottled background. A file can carry an enormous pixel count and fail on exactly that. Note the unit it uses when it does state a size: inches, not pixels.
Why DPI is meaningless until the print size is fixed
DPI is not a property of your pixels. It is a number in the file's header saying how many of those pixels to lay down in each printed inch. Change it with resampling switched off and not one pixel in the image changes — you have edited a label.
The only figure with a physical meaning is pixels divided by printed inches, which is why this post carries both: a square measured in inches, quoted from the manual above, and 2×2 in (51×51 mm) · 600×600 to 1200×1200 pixels, square for the file. Turn resampling on and it is no longer a label being edited — the image is rebuilt at a new pixel count, which does change compliance, in either direction. A tool that "sets the dpi" without touching the pixel dimensions changes nothing.
The file size floor, and the habits that trip it
The file has a weight range too — 54 KB to 10 MB — and the floor, not the ceiling, is the end that catches people. A minimum file size is a proxy for detail: at the pixel dimensions the spec requires, a JPEG can only get that small by having its detail compressed out, and what comes back is the "grain-like particles" the manual says "may not be acceptable".
Three habits produce an under-weight or under-sized file from a photo that started out fine:
- Over-compressing on purpose. Hearing that uploads reject large files, people run the image through a compressor at its heaviest setting and land under the floor. Compress toward the middle of the range, never the bottom.
- Sharing it first. A messaging app, an email client's "small / medium / large" prompt, or an "optimise for web" export re-encodes at a reduced pixel count without announcing it. The file on your laptop is not the file that left your phone.
- Screenshotting it. A screenshot replaces the photograph's pixels with your screen's pixels, keeping whatever size it happened to be displayed at. A screenshot of a compliant photo is rarely a compliant photo.
The container those pixels arrive in is a separate problem — JPEG is what the upload expects, and an iPhone HEIC or a PNG renamed rather than converted is still not that:
My phone shoots 4032 × 3024 and it was still rejected
The pixel count is a red herring here. A twelve-megapixel phone camera produces a four-by-three rectangle — around 4032 × 3024 — with more pixels than the spec asks for and the wrong shape. The requirement is square, and a rectangle is not a square at any resolution.
Cropping it square is the right move, and where the second failure appears. Cutting a square out of a four-by-three frame throws the sides away, and what is left still has to satisfy 50–69% of the photo's height (1 to 1⅜ in / 25–35 mm) and 56–69% up from the bottom of the photo. A square chosen by eye almost always leaves the head too small — the arm's-length selfie problem — or the eye line too low.
So the order matters: decide the crop from the head and the eyes, then confirm the result lands inside 2×2 in (51×51 mm) · 600×600 to 1200×1200 pixels, square. Doing it the other way round produces a file that passes every dimension check and fails the composition one.
What upscaling does, and the headroom it costs you
If a photo sits below the floor, the obvious move is to enlarge it — and that is what turns a fixable photo into an unfixable one. Upscaling invents pixels: it cannot recover detail the camera never recorded, only interpolate between the pixels you have, producing precisely the softness and grain the quality rules name.
The second cost is the one nobody publishes. Spare resolution is not decoration — it is the margin that lets an adjudicator rescue a photo instead of returning it. Where the facial image is too small in the frame, 8 FAM 402.1 says: "If the photograph has sufficient resolution, the photograph may be re-sized in the TDIS image review module." Where it is too large, the same escape hatch applies if "there is sufficient room to crop the photograph appropriately".
So shoot at full resolution and crop down rather than shoot small and scale up: an upscaled file has the pixel count without the headroom. Let distance rather than proximity do the framing, too — filling the frame by bringing the camera close to the face has its own cost. The manual warns that "in some cases, a larger facial image size may cause fish-eye distortion", the nose exaggerated and the edges of the face diminished, which is a rejection in its own right.
The UK publishes a pixel floor, and no ceiling at all
If you arrived here for a UK passport photo, the rule is published as a different kind of number. GOV.UK's digital photo rules give a floor and nothing above it: the photo must be "at least 600 pixels wide and 750 pixels tall" and "at least 50KB and no more than 10MB". There is no maximum pixel size on that page and no stated aspect ratio, so a minimum is not a shape — a file can clear the UK floor and sit inside the US band at the same time.
What differs is the picture being asked for, not the arithmetic. GOV.UK tells online applicants to "include your head, shoulders and upper body. Do not crop your photo - it will be done for you." A US photo is a square you crop yourself to 50–69% of the photo's height (1 to 1⅜ in / 25–35 mm) and 56–69% up from the bottom of the photo; a UK digital one is wider, and you are told to leave it alone.
One disclosure, because that instruction is one our own product does not clear, along with the requirement that a digital photo be "unaltered by computer software". What we produce is cropped and has its background replaced, so we do not claim our UK photo for the fully online UK application. It is built for the UK printed route, which you print and post.
Measure the pixels on the photo you already have
Pixel dimensions are the one item here you can read off a file yourself; your phone and your computer both show them in the file's info panel. What the same upload measures alongside them is harder to judge — the head band, the eye line, the background, the sharpness. These are the checks a US passport photo is held to here:
- Exactly one face, looking straight ahead
- Head sized 50–69% of frame
- Eyes at the required height (56–69% from the bottom)
- No head tilt
- Background replaced with pure white
- Even, uniform background
- Sharp focus — no blur
- Natural exposure — not too dark or bright
The check is free, needs no account, and reports every rule as a pass or a fail with the measured value beside it — so a photo inside 2×2 in (51×51 mm) · 600×600 to 1200×1200 pixels, square but outside 50–69% of the photo's height (1 to 1⅜ in / 25–35 mm) is told exactly that, rather than left to guess which of the two an upload objected to.
If it passes, the finished file is 600×600 JPEG · 2×2 in print sheet · pure white background, with a print-ready 4×6 sheet (6 copies) for the paper form. If it fails, the report names the rule, which is usually enough to know whether the photo is a crop away or a retake away.
Measure my photo against every rule
See your finished photo and its full check report free · download it for $16.99 · refunded if the government rejects it