Extract data from passports
Unstract reads passports, driver's licenses and ID cards from scans and phone photos, checks the machine-readable zone against the printed page, and sends each customer's identity data to your onboarding or KYC system.
seen original at branch 14 — JR
Specimen data page · fictional holder
{ "document_type": "Passport", "issuing_country": "UTO", "passport_number": "X74209163", "surname": "Whitfield", "given_names": "Elena Rose", "nationality": "Utopian", "date_of_birth": "14 Mar 1988", "date_of_expiry": "08 Jun 2031", "mrz_check_digits": "passed", "confidence": 0.97}
Trusted by lending and onboarding teams around the globe
Customers send their ID the way their phone or branch captured it
A clean scan from a branch desk is rare. Most IDs reach the onboarding queue as a photo taken on a kitchen counter, at an angle, under a lamp.
Phone photos with glare on the laminate and a thumb over one corner
One date of birth, three labels across a passport, a license and an ID card
Marriage certificates filled in by hand to explain a name change
Four identity documents in one PDF from the account-opening portal
License front and back sent as two separate images
Onboarding staff type names, numbers and dates from each ID into the KYC record and compare them with the application by eye, so new accounts wait in a review queue and customers abandon the signup.
Unstract reads the ID in whatever condition it arrives, extracts the holder's details, checks the MRZ, the expiry date and the application, and posts the result to your onboarding or KYC system.
From an ID photo to a matched applicant record
Read the ID, extract the holder's details, then run customer onboarding around them. Each product works alone, and the three together take an application to screening without anyone retyping a passport number.
Read passport data pages and both MRZ lines as clean text
LLMWhisperer turns a scanned or photographed passport into text that keeps each label beside its value, so a date of issue is never taken for a date of birth. It straightens tilted phone photos, reads through glare and noise, and returns the two machine-readable lines exactly as printed, filler characters included, so MRZ OCR has every character it needs to verify.
Send that text to your own LLM, or make it the first step of customer onboarding.
Learn more about LLMWhispererTurn passports and other IDs into one identity record
Name the fields your KYC record needs, and Document Studio returns them in one JSON shape whether the ID is a passport, a state driver's license or a national ID card. The printed page and the MRZ are read separately and compared, so a misread digit or an altered date shows up as a disagreement instead of slipping into the file. An expired document or a birth date that differs from the application comes back flagged with the reason.
Developers call it as a passport OCR API from a signup flow or a mobile app, operations teams run it as an ETL pipeline over the onboarding uploads folder, or it becomes one step of an orchestrated onboarding run.
Learn more about Document StudioRun customer onboarding from the application to screening
Describe how your team verifies a customer's passport or driver's license in plain English, and Flow's agents build the workflow. Each upload is sorted by type, the ID is read and matched against the application, proof of address is checked for age, and matched files go on to sanctions screening such as LSEG World-Check One or LexisNexis Bridger Insight XG, then to your CRM or core banking. You approve the workflow once, and every application runs through it the same way.
You choose when a person steps in: a name or date of birth that does not match, an expired ID, proof of address older than 90 days, or business owners that do not add up. Only those files stop, and a KYC analyst sees the documents side by side with the mismatch marked.
Learn more about FlowRead your customers' IDs today. Automate onboarding when you're ready.
Start for free and run your own passports and licenses through parsing and extraction now. Matching, screening hand-off and posting to your onboarding or KYC system start with a consultation.
Passports, licenses and the identity papers that come with them
Documents
- Passports
- Passport cards
- Driver's licenses
- Learner's permits
- ID cards
- National identity cards
- Employment authorization documents
- Permanent resident cards
- Social security cards
- Visas and entry stamps
- Voter registration cards
- Military ID cards
- Marriage certificates
- Name change court orders
- Birth certificates
- Proof of address (utility bills)
Fields captured
For the teams that open accounts and answer to examiners
Branch and online applications move from upload to an opened account the same day, with the customer's ID details already in the record and no one copying a passport number off a photocopy.
A signup flow reads the passport or license a customer photographs in the app, so a new user gets an answer in minutes and drop-off falls.
Analysts work only the files that fail a check, see the ID beside the application, and keep a record of what was read, flagged and cleared for every customer.
New account paperwork for individuals, joint holders and trusts is read the same way, including the marriage certificate that explains why the name on the passport differs from the one on the application.
ID details matched before the account is opened
Guarding the data an identity thief wants most
A passport or driver's license puts a person's full name, date of birth, document number and photo on one page. Those are the details a fraudster needs to open accounts, take out loans or pass someone else's verification, and unlike a card number they cannot be reissued overnight.
- Encrypted in transit and at rest.
- Customer IDs are never used to train models.
- Access by role and onboarding queue, with every corrected field and cleared flag logged against the customer record for your examiners.
- You set retention: ID images are deleted when your BSA and privacy policies say they can go, and the extracted record stays.
- Self-hosted deployment for banks and credit unions that keep customer identity data inside their own environment.
What onboarding and KYC teams ask about reading IDs
Send the passport image to Unstract instead of to a person. Customers upload a photo or scan through your portal, LLMWhisperer reads the data page and the MRZ, and Document Studio returns the name, number, dates and nationality as JSON. Run it from an API call, a watched folder or a full onboarding workflow in Flow.
Yes. Document Studio deploys an endpoint per document type: POST the passport image and get the holder's record back as JSON, with the MRZ check result included. The same passport OCR API serves driver's licenses and ID cards, so one integration covers every ID a customer might send.
Upload the file, list the fields you want, and run it. Start for free on your own customers' passports, then move the same extraction into an ETL pipeline or an API once the fields look right.
The two machine-readable lines repeat the passport number, birth date and expiry with a check digit after each. Unstract reads both lines, recomputes the check digits and compares the MRZ with the printed fields, so a misread character, a damaged line or an edited date is flagged instead of accepted.
Yes. Driver's licenses, ID cards, employment authorization documents, social security cards, voter registration cards and marriage certificates are read with the same approach, along with the proof of address and business formation documents that arrive in the same application.
No. Unstract is software, not scanning hardware. It reads the images you already receive: a flatbed scan from a branch, a passport scanner at a desk, or a phone photo from your app. If your branches already use a drivers license scanner, send its images straight in.
Passports that follow the ICAO layout share the same data page fields and MRZ format, and Unstract reads the printed fields by meaning, so a new country's design needs no template. Non-Latin names are read alongside the Latin transliteration in the MRZ.
The file stops for a KYC analyst with the ID and the application side by side and the differing field marked. The analyst corrects, requests a new document or clears it, and the decision is recorded against the customer.
Yes. The expiry date is extracted from the printed page and the MRZ, compared with the application date, and an expired or soon-to-expire ID is flagged before the account is opened.
Unstract is SOC 2, ISO 27001, GDPR and HIPAA compliant. ID images are encrypted in transit and at rest, never used to train models, and deleted on the schedule you set. Banks that keep identity data in-house can self-host; see the Trust Center.
Bring a day's worth of customer IDs to the consultation
Glare-covered passport photos, license fronts and backs, an employment authorization card and the file where the names don't match. We read them live and show the fields, the MRZ check and every flag.