How PhotoCheck verifies your photo

The 8 criteria we evaluate

Every check returns a structured report scoring the photo against eight criteria. Each criterion is marked pass, fail, or uncertain, with the specific rule clause applied and a recommendation if it fails.

The criterion scopes are strict. Head size belongs to dimensions, not head position; glasses belong to eyes, not coverage. We enforce these scopes so the output of one check is directly comparable to another.

How the analysis works

Photo analysis runs on a leading third-party vision model, accessed with zero-retention configured: the provider does not retain or train on the images or prompts we send. The Prep tier adds an image model for background replacement and lighting correction, and uses facial-landmark detection to drive the submission crop.

The analysis prompt is restrictive. The model is instructed to stay observation-based, prefer uncertainwhen a measurement cannot be made from the image alone, and never describe the person’s race, age, gender, or perceived nationality. During Prep the subject is never modified — face, hair, clothing, expression, head angle, and accessories are all preserved exactly; only the background and lighting may change.

Sources we cite

Each country’s rule set is encoded from the official issuing authority and re-verified periodically. The model is given this rule data verbatim as part of every check.

Government photo rules change a few times a year. Where our copy and the linked official source disagree, the official source is authoritative. Per-country pages also list this date at the top.

Pricing and rate limits

• Free check. No account, no email. Rate-limited to a small daily quota per IP address to keep abuse cheap; legitimate retakes (running the same photo a few times after small adjustments) always work.
• $1 Prep. One-time. Produces a submission-ready JPG resized and recolored to the destination country specification, plus a 4×6 inch print sheet at 300 DPI with four copies arranged for cutting. For comparison, retail passport-photo services typically charge $15–$30 for a single submission file.

What we claim, and what we do not

Known limits of automated analysis

• Exact measurements in millimeters (head height, chin-to-crown, background RGB) cannot be inferred from a digital image without a known scale reference. Where this matters we mark the criterion as uncertain rather than guessing.
• Color calibration shifts between digital and printed forms. Our analysis is on the digital file you upload; a printer may shift the background away from the target white or grey.
• Some consulates enforce unwritten preferences not captured in the published guidance. We encode what is published; we cannot encode every regional variation.
• The image model used in Prep for background replacement very occasionally fails to preserve a subject perfectly (rare hair artifacts, faint lighting changes). We mitigate with strict preservation rules and a facial-landmark-driven crop, but it is not perfect; review the output before submitting.

Reproducibility

For a given photo, country, and rule snapshot, the same model generally returns the same report. The model is non-deterministic (temperature is left at the model default), so wording of the observation field can vary between runs even when the pass/fail/uncertain verdict is stable. If you suspect a clear technical error in a verdict, the check is free to re-run.