Guides & How-Tos

Why Passport and Visa Photo Apps Reject Your File

20 min read
Why passport photo uploads fail

You crop the file to the exact millimetre the page asked for, upload it, and three days later the application comes back with a rejection notice that tells you almost nothing. No pixel dimensions, no byte count, no named rule. Just a photo that failed, and you starting over from a blank canvas. That loop is the single most common complaint I hear about hosted passport photo editors, and it exists because a rectangle is the easy half of the job.

By the end of this you will know the checks a digital submission passes through before any person opens it, which of those checks you can measure yourself in seconds, and which failures only a retake can fix. Most of it happens in a tab that never contacts a server.

Why a Perfect Rectangle Still Gets Rejected

Your file gets measured automatically. A human then checks the three things a ruler cannot judge: how tall the head sits inside the frame, where the eyes land, and whether the background is clean. A photo can be 35 x 45 mm to the pixel and still fail, because the crop only fixes the canvas.

Head height is measured from crown to chin, and it fails in opposite directions. Too small leaves dead space above the head and along both sides, which reads as a portrait shot from too far back. Too large pushes the crown or the chin against the frame edge, and reviewers reject that as cropping. The State Department is blunt about the fix too. Its guidance warns against altering a passport photo with software, phone apps, filters, or artificial intelligence, which rules out most hosted editors that promise to “fix” a background after the fact.

That warning matters more than it first appears. Any app that reskins the wall behind your head is doing exactly the kind of edit the official guidance tells you to avoid, and it produces a file that can pass every automated check and still fail a reviewer who compares it against you in person. If you want to check the measurable parts of a file before you upload it, crop and check a passport photo against the official frame rules without uploading anything.

Three things pass the crop and still fail review:

  • Head height outside the office’s millimetre band
  • Eyes below the expected line, or a face turned even slightly off axis
  • A shadow, colour cast, or texture in what should be a flat backdrop

The Automated Screen That Runs Before a Human Sees It

A digital submission meets a script before it meets a person. If the file fails that script, it never reaches a queue, which is exactly why the rejection notice is so vague. The gate is boring and completely deterministic, and every one of its checks is something you can verify locally. The trouble is that most people only find out which check failed after the resubmission window has closed, so running the checks yourself first is the difference between fixing a file and starting a new application.

The Checks the Portal Runs First

The US Department of State lists the requirements for visa and Diversity Lottery images in numeric terms, because a script reads them. The same list, in the order you would verify it:

  1. Format. JPEG for visa and DV submissions. Online passport renewal accepts JPG, PNG, and HEIF as well.1
  2. Aspect ratio. Square, 1:1, no letterboxing and no visible border.
  3. Pixel dimensions. Between 600 x 600 and 1,200 x 1,200 pixels for visa images. A 2 x 2 inch US passport frame lands at 600 x 600 at 300 DPI.
  4. Byte cap. 54 KB to 10 MB for online passport renewal, and 240 kB maximum for visa and DV submissions. The lower end matters as much as the upper.
  5. Colour. 24-bit sRGB, checked against the embedded profile.
  6. Compression ratio. 20:1 maximum, and the face has to survive the squeeze.1

That sixth item is where a lot of people lose. Hitting a byte cap by cranking quality down produces a small, technically valid file with a mushy face, and facial detail is part of the requirement rather than a nice-to-have. A script cannot judge that, so it passes a file that a reviewer rejects an hour later, which is the worst possible outcome.

Flow diagram of the six automated checks a passport or visa photo upload passes through, in order: format, aspect ratio, pixel dimensions, byte cap, colour profile, and compression ratio
The portal's script walks the same six checks in the same order every time, and each one is a number you can read off your own file before uploading.

What Colour Profile and Compression Ratio Have to Do With It

A JPEG tagged AdobeRGB or Display P3 shifts skin tones, and the rejection lands on you even though the file looks correct on the monitor you made it on. Broad-gamut displays render a wide-gamut file as intended, so the error is invisible until a portal that expects sRGB reads the numbers. If you work across several colour spaces and need to know what you actually have, convert colours between sRGB, AdobeRGB, and Display P3 and see the numeric difference before the upload rather than after.

Compression works the same way in reverse. Saving at quality 40 to land under 240 kB is a reasonable move, and saving at quality 15 to land under 240 kB is a different decision with a different face. If the original is only 600 x 600, there is usually no need to compress at all.

Head Height, Eye Line, and the Percentage Rule

The US rule reads like two different rules because it is two different rules. Stated in millimetres, the head must sit 25 to 35 mm from chin to crown inside a 2 x 2 inch frame. Stated as a proportion, the same requirement says the head must fill 50 to 69 percent of the image height. Offices quote whichever form suits them, and both land on the same measurement.2

A head that is too small fails because the proportion sits under 50 percent. A head that is too large fails because the crown clips the top edge or the chin drops past the bottom. The failure is not “close enough” in either direction, and the band is wide enough to give you real latitude once you know where it sits.

The EU Visa Code expresses the same idea in percentage terms for everyone. The face should cover roughly 70 to 80 percent of the photo height, and inside a 35 x 45 mm frame that becomes a 32 to 36 mm band. Wider head requirement than the US, smaller frame, and therefore more of the frame taken up by the face. Comparing the two side by side is the fastest way to understand why one cropped photo cannot satisfy both.3

Eye position gets treated differently by different offices, and some care more than others. The US guidance describes eye position relative to the frame’s vertical centre, while European practice tends to describe the whole face placement instead, so two people can each be technically right and still produce different crops. A dashed eye-line guide in an editor is a measuring aid. It is not an approval, and a tool that draws the line for you still cannot tell you the office’s exact tolerance.2

Glasses, Head Coverings, and How Old the Photo Must Be

Two of the most common rejections have nothing to do with geometry, and no crop setting can fix either one. The US guidance is explicit: you are not wearing glasses, and you cannot wear a hat or head covering except for religious or medical purposes. Glare on a lens, a faint frame edge across the temple, and a reflection of the photographer’s phone all disappear in a crop and reappear in person, which is why a photo with clear glasses can be refused by an officer who was handed nothing but a correct rectangle.

Head coverings follow a narrower path. A covering worn daily for religious reasons is permitted where the guidance allows it, and it has to be one solid colour with no patterns or small holes, since a patterned scarf throws exactly the background texture the frame forbids. If the covering sits back far enough to expose the hairline, that also settles it. Every other case is a retake, and no editor offers a legitimate way to produce one.

Age is the third one, and it is the easiest to get wrong because nothing about the file reveals it. US passport and visa photos must have been taken within the last six months, which is about 180 days and considerably less than a holiday. A photo that was perfect when you took it stays perfect on disk forever, and the portal has no way to tell. Nobody stamps a date on a JPEG, so the check happens entirely in your head at the moment of upload. Reusing an older photo is one of the few rejections that a compliant, correctly cropped, technically flawless file cannot avoid.

The practical ordering follows from all three. Settle what you are wearing and how old the shot is before you open a cropper, because those two decide whether a crop is worth doing at all.

Background Variance Is the One Failure You Can Measure

Official background guidance is written for a human reviewer, which makes it miserable to self-check. “Plain white or off-white, free of shadows, textures, or objects” describes what a good photo looks like, not what a number looks like. Squinting at your own image in a photo editor tells you almost nothing, because your eye adapts to the screen brightness and the surrounding room.

The repeatable version samples the luminance of the output frame outside a padded ellipse around the estimated head, then takes the variance of those samples. A flat wall produces a low number. A wall with a shadow gradient, a colour cast, a wall texture, or a stray shadow from your own head produces a high one. In the cropper I use, a variance above 400 reports as a fail with the measured value shown next to the maximum, so you can see the gap instead of guessing.

Diagram of a portrait frame with a padded ellipse excluded from sampling, sample points across the background, and a variance scale where a flat white wall sits low and a wall with a shadow gradient reads above the threshold of 400
Background variance samples the frame outside a padded ellipse around the head, and anything past 400 means the wall, not the crop, is the problem.

The detail that matters most is what happens next. A failed background row stays visible on the compliance board and does not block your download, because the fix is a retake. Smoothing the skin or pasting in a cleaner backdrop is precisely the software edit the State Department tells you to avoid, and it produces a file that can fail a manual review after passing the automated one.

Distance from the wall and even lighting move that number. A lamp above and behind the camera casts a shadow you cannot see in a mirror selfie, and moving to a metre out with the light in front of you is the whole remedy rather than a retouching step.

Searching for the official hex value of passport-photo white comes up empty, and that absence is the actual answer. No authority publishes a numeric colour target for a background, which means “plain white or off-white” can only ever be judged against a reference print and an inspector’s memory. A wall painted bright white under warm evening light reads as cream in the file, and cream passes in the UK and fails in the US. Nothing in the number is wrong, because there is no number to match.

Picking the Right Frame Before You Crop Anything

The country decides the frame, and nothing else does. Guessing is how people end up resubmitting to a UK portal with a US 2 x 2 inch crop that looks fine and fails anyway, which costs a week of waiting for a rule that takes ten seconds to check. Here is what the standard presets actually store:

PresetFrameHead bandBackgroundOutput pixels
UK passport (printed)35 x 45 mm29–34 mmCream or light grey4827 x 1063
US passport2 x 2 in25–35 mmWhite or off-white600 x 600
Schengen / ICAO35 x 45 mm32–36 mmPlain light413 x 531
Indian passport35 x 45 mmno stored limitWhite630 x 810
Canadian passport50 x 70 mm31–36 mmWhite or light-coloured591 x 827
Australian passport35 x 45 mm32–36 mmWhite or light grey5413 x 531

Canada is the outlier and it catches people out. The Canadian passport photo specification uses a 50 x 70 mm frame with a 31 to 36 mm head, which is taller and narrower than the European standard everyone assumes as the default, so a crop that worked for a Schengen application will not survive the same upload north of the border.

The pixels are arithmetic rather than preference. Millimetres convert at the preset’s DPI with px = mm / 25.4 * dpi, rounded to whole pixels, so a 35 x 45 mm frame at 300 DPI produces 413 x 531. A row reading “no stored limit” means the preset data holds no band for that country, not that some office has blessed your photo. Australia permits a wider range than the 32 to 36 mm band the preset uses, and the UK digital flow asks for at least 600 x 750 px and crops during the application rather than taking a finished file.6

35 x 45 mm appears in five of six rows for a good reason. It traces back to ICAO Doc 9303 Part 3, the specification for machine-readable travel documents that most countries issue against, so a photo compliant with it moves between passports more easily than a photo tuned to one office. If you would rather check the frame math against other browser-based tools that process everything locally, like CapyToolkit’s own free utilities than against a PDF, that is a reasonable way to work.

Read the pixel column twice, because it holds a trap. The UK row produces 827 x 1063 and the Schengen row produces 413 x 531, and both describe the same 35 x 45 mm piece of paper. That difference is DPI, and a higher DPI is not a better photo; it is the same photo carrying more bytes. You get four times the pixels, roughly four times the file size, and exactly zero extra facial detail, because the head fills the same fraction of the frame either way. Under a 240 kB cap the 600 DPI version simply gets compressed four times harder for no visible gain.

Side-by-side comparison of the same 35 by 45 millimetre passport frame rendered at 413 by 531 pixels and at 827 by 1063 pixels, showing identical head framing with quadruple byte cost
Both frames are the same 35 x 45 mm of paper; the 827 x 1063 version carries four times the pixels, roughly four times the file size, and no extra facial detail.

So pick the smallest number in the column that still clears your portal’s minimum. Between 600 x 600 and 1,200 x 1,200 for a US visa image, take the bottom of the range unless something forces your hand, and note that 413 x 531 satisfies a physical print while failing an online square requirement outright.

Digital File Limits and Printed Sizes Are Different Problems

A reader applying on paper and a reader applying online are solving two different jobs, and official pages do not always make it obvious which rule belongs to which. A byte cap has no meaning for paper. A physical head-height band has no meaning for a portal that never prints anything.

When the Portal Wants Pixels

Downloading at the preset’s exact width and height gives you a file that is the frame itself, not a larger canvas with a white border painted around it. Border padding is a common upload mistake, and it survives a casual look because the border reads as part of the white background.

The measured JPEG size on a compliance board exists to be compared against the portal’s byte limit. It is information, not a verdict, and it changes with the quality setting you pick rather than with the document you are applying for. A 220 kB file at quality 92 and a 220 kB file at quality 70 both satisfy the cap, and only one of them survives a manual review.

When the Office Wants Paper

Print sheets place 4 or 6 copies at true physical size on A4 or 4 x 6 inch paper. Where the requested count does not fit the chosen paper at that size, a good tool keeps the largest fitting count and tells you it reduced, rather than quietly shrinking the photos into something that is no longer 35 x 45 mm. Silent scaling is how a “compliant” print turns out to be 30 x 38 mm on a ruler.

UK paper applications want two identical prints kept loose, not attached to the form, and the GOV.UK printed photo guidance spells out the 35 x 45 mm size, the 29 to 34 mm head band, and the cream or light grey background. Take a ruler to the print shop and state the intended physical size. A printer that defaults to “fit to page” is the most common cause of a rejected paper application.

Shrinking a File Without Damaging the Face

The tempting fix for an oversized file is to re-save it smaller, and the way most editors do that is exactly what trips the compression-ratio and facial-detail checks. Quality sliders make the byte count drop on demand, which feels like control, but nobody checks whether the eyes still have detail afterwards.

Going the other direction is more common. A modern phone photo is far larger than a 600 x 600 square needs, so the right order is crop to the frame first, then look at the byte count. If the file is still over the cap after cropping, shrink a photo to a specific file size without a round trip to a server and verify the result at 100 percent. If the face looks blocky or smeared when you zoom in, the compression has already gone too far, and no amount of file-size compliance rescues it.

There is one more mistake worth naming, because it looks like the opposite of shrinking. Upscaling a small file to hit the minimum pixel count produces a photo with no real detail behind the upscale, and a reviewer comparing it against a live person at the counter can tell. If your source is below 600 x 600, the answer is a better source photo, not a larger one. Every phone from the last decade overshoots the requirement, so a small file almost always means a cropped crop.

Check the bytes too, not just the dimensions. A file exported from a design tool often carries a wide-gamut profile and several hundred kilobytes of editing metadata, neither of which a portal asked for and both of which can trip the checks above. Fix the profile in your export dialog before the crop, and strip embedded metadata before the file leaves your machine.

Keeping a Biometric Photo Off Someone Else’s Server

A passport photo is a biometric document, and that is a different risk class from a tax receipt or a warranty claim. It shows your face at a known ratio, beside your name, in the exact proportions an identity document requires, which means the file is useful to anyone building a face index even if it never becomes a passport. Most hosted editors are not malicious about this. They are just optimising for a queue, a render farm, and a CDN, and your face rides through all three.

What client-side processing means here is narrower and more testable than a privacy promise. Your browser decodes the image, pans and scales it, measures the head against the guide, encodes a JPEG, and renders the result. Every one of those steps is a function call on bytes that never left the address bar. No policy page to read, no retention window to reason about.

Two honest caveats, because a claim of zero network activity that turns out to have an exception is worse than no claim. The optional face-detection guide does fetch its model assets, and that request exists precisely because the guide needs landmarks to place itself; what it carries is the model, not your photo pixels. And camera permission starts only when you open the camera panel, so declining it leaves the file-upload path working exactly as before. Closing the tab releases the in-memory image and the generated output.

One practical gap worth knowing before you start. HEIC and HEIF from an iPhone are not decoded in a browser tab, so convert to JPEG on the device first.7 A photo that never opens is the least recoverable failure in the whole workflow, and it costs one conversion step in the camera roll to avoid it.

Before you spend money on prints, check this list:

  • Output pixels match the portal’s range exactly, and are square if it wants square
  • Measured head height sits inside the band for the country you are applying to
  • Background variance is comfortably under the threshold
  • Byte count is inside the cap, and the face still has detail at 100 percent
  • Colour profile is sRGB, not a wide-gamut profile from a design workflow
  • Printed copies are loose, at true size, and measured with a ruler

Every one of those is checkable before you walk out of a print shop, which takes about five minutes of your time and saves a trip back. The one failure none of them catch is scale in the negative direction, and that one is a retake rather than a setting.

Sources
  1. 1.

    U.S. Department of State, “Digital Image Requirements,” travel.state.gov, accessed September 2026. https://travel.state.gov/content/travel/en/us-visas/visa-information-resources/photos/digital-image-requirements.html

  2. 2.

    U.S. Department of State, “Photo Composition Template,” travel.state.gov, accessed September 2026. https://travel.state.gov/content/travel/en/us-visas/visa-information-resources/photos/photo-composition-template.html

  3. 3.

    European Parliament and Council of the European Union, “Regulation (EC) No 810/2009 establishing a Community Code on Visas (Visa Code),” EUR-Lex, September 2009. https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32009R0810

  4. 4.

    GOV.UK, “Get a passport photo: Printed photos,” www.gov.uk, accessed September 2026. https://www.gov.uk/photos-for-passports/photo-requirements

  5. 5.

    Australian Passport Office, “Passport photos,” www.passports.gov.au, accessed September 2026. https://www.passports.gov.au/help/passport-photos

  6. 6.

    GOV.UK, “Get a passport photo: Digital photos,” www.gov.uk, accessed September 2026. https://www.gov.uk/photos-for-passports

  7. 7.

    Can I Use, “HEIF/HEIC image format,” caniuse.com, accessed September 2026. https://caniuse.com/heif

More in Guides & How-Tos