ID Card OCR API for National Identity Cards in 25+ Countries

The API reads national identity cards from 25+ countries across Asia, Europe, Africa and the Americas — including Thailand, China, Indonesia, India, South Korea, the Philippines, Germany, France, Italy and more. Every supported document has its own field reference, linked from the supported-countries grid on this page. The list grows continuously: if you need a country we don’t list yet, request it and we add most new document types in 1–2 days.

Because national ID cards genuinely carry different information. A Thai ID prints religion and a Buddhist-calendar date of birth; a German ID prints height and eye color; a Chinese ID prints ethnicity and issuing authority. Rather than flatten everything into a lowest-common-denominator schema, the API returns the fields each country’s document actually contains, using that document’s own field names. You get a core set — ID number, name, date of birth, address, validity dates — plus the country-specific extras. Check the country reference for the exact keys you’ll receive.

Native-script recognition is core, not an afterthought. The engine is trained on Thai, Chinese, Arabic and other non-Latin scripts, and returns the name and address in the card’s original script exactly as printed. When the card also prints a romanized or Latin form, the API returns that alongside the native-script value, so you have both a local and a Latin handle for the same person. This is what makes the API usable for multilingual ID OCR in global KYC, where forcing everything into Latin would corrupt names.

documentType is country-specific for ID cards — for example th_id_card, de_id_card, cn_id_card — and it tells the engine which national format to read. You pass it with the image in the same request. The complete list of values is in the supported-countries grid on this page; each entry links to that document’s reference. If you don’t know the country in advance, capture it from your own onboarding flow (the user’s selected country) and map it to the matching documentType.

Yes — the engine is built for real-world capture, not just clean scans. Detection and preprocessing handle glare, rotation, perspective skew and moderate blur from handheld phone photos before OCR runs. Image quality still matters: better-lit, in-focus, fully-in-frame photos yield better extraction. For onboarding flows, guiding users to capture the card flat and well-lit measurably improves results, but you don’t need a controlled scanning environment.

The API is designed for the data-extraction step of KYC and AML workflows: it turns a national ID card into structured fields — ID number, name, date of birth, validity — that you can match against an application and feed into screening. It is an OCR and data-extraction tool, not a full identity-verification or decisioning service: liveness, document authenticity scoring and sanctions screening are separate concerns you combine around it. Many teams use it as the extraction layer underneath their existing KYC stack.

Dates are normalized to ISO 8601 (YYYY-MM-DD), including conversion from local calendars — a Thai card’s Buddhist-year date of birth is returned both in its original form and as a Gregorian ISO date. ID numbers are returned as printed. Exact normalization behavior per field is documented in each country’s reference.

Images are sent over TLS and processed to return the structured JSON response. We treat identity documents as sensitive data; for data handling, retention and a data-processing agreement, see our privacy policy or contact us. If your jurisdiction has specific residency or retention requirements for ID data, raise them with us before you integrate so we can confirm the right setup.

Sign up for a free trial, get an API key, and POST your first ID image with the matching documentType — you’ll have structured JSON back in seconds. Pricing is usage-based with a free tier to start; see the pricing page for current tiers. Most teams begin by running the IDs they already collect through the API to measure extraction on their real traffic before rolling it into production.