Germany passport OCR API

Recognising a Germany passport

A Germany passport follows the ICAO 9303 standard: a two-line, 44-character machine-readable zone (format TD3) printed at the foot of the data page, below the human-readable visual inspection zone. Its first five characters are P<D<< – the document code, then the issuing stateas D, a code ICAO 9303 keeps for Germany instead of its ISO 3166-1 alpha-3 code DEU. The API reads both zones, cross-checks the MRZ check digits, and returns one response shape with every key present.

Germany is DE in an alpha-2 list and DEU in ISO 3166-1 alpha-3, and its zone prints neither: ICAO Doc 9303 gives it the code D, filled out with < to three characters (D<<). A parser that maps zone codes to countries through the ISO table alone will not find Germany.

What is particular to the Germany passport

Issued by the passport authorities (Passbehörden) the federal states designate, and German missions abroad.

• The zone writes Germany as D, not DEU: the Passport Act prescribes the abbreviation D, and ICAO Doc 9303 part 3 lists D as Germany's code.
• By law the zone's document code is PP for the ordinary passport, PO for the service passport and PD for the diplomatic passport.
• The sex field is F, M, or < for a holder of another sex.
• A passport is valid for 10 years, or 6 years for a holder under 24, and its validity cannot be extended.

Sources:

• Passgesetz § 4 – the passport's contents
• Passgesetz § 5 – validity
• Passgesetz § 19 – the passport authorities
• ICAO Doc 9303 part 3 (PDF)

Machine-readable zones issued by Germany

9303 defines three zone layouts, and the API reads all three. Below is one invented holder – she is nobody, and paradeigma is the Greek word for example – in each layout, with D as the issuing state. Every check digit is computed from the characters beside it rather than typed in, so each zone here validates.

P<D<<PARADEIGMA<<ELENI<SOFIA<<<<<<<<<<<<<<<<
AM73045184D<<9403084F3203101PN48291630<<<<72

I<D<<PARADEIGMA<<ELENI<SOFIA<<<<<<<<
TR61850420GRC9403084F2911043<<<<<<<2

IDD<<AK472913<4PN48291630<<<<<
9403084F3405204GRC<<<<<<<<<<<9
PARADEIGMA<<ELENI<SOFIA<<<<<<<

No check digit covers the issuing state: on all three layouts the arithmetic runs over the document number, the dates and the optional data, never over the three characters that say who issued the document. A zone whose digits all pass has not proved it came from Germany – the visual zone and the document's own security features do that.

Paste an MRZ into the MRZ parser to validate its check digits, or build one with the MRZ generator.

Fields you get back

Every recognised German passport returns the same structured fields. A value that could not be read is null, never a missing key.

• Surname and given names
• Date of birth and sex
• Document number and personal number
• Nationality and issuing state (D)
• Issue and expiry dates, days remaining
• The MRZ verdict, and the lines it was read from

See every field in the document field explorer.

Make the call

Run this against the public sandbox – no registration, real recognition, free for a limited number of documents and 10 requests per hour per IP. Swap the sandbox key for your own live key to keep recognising German passports.

curl -X POST https://api.doc.cheap/v1/scans \
  -H "Authorization: Bearer sk_sandbox_public" \
  -H "Content-Type: application/json" \
  -d "{\"image\": \"$(base64 < document.jpg | tr -d '\n')\"}"

Where this is documented

This page covers what is specific to Germany. The four pages below carry everything that is not.

• Recognize a passport

  The call above with a real document, in curl, JavaScript and Python, and every key of the answer read out.

• The MRZ and the visual zone

  Why a document prints the same facts twice, which zone to believe when they disagree, and what the check digits above can and cannot prove.

• Field languages and scripts

  The zone is Latin transliteration; the visual zone is not. Which reading carries the original script, and how a language identifier resolves to a language.

• Errors

  Every error code the API can answer with, in one table: what causes it, whether a retry can succeed, and whether the call was billed.