Work/AI
Cisevo Document AILive product

Reads documents from imperfect photos and verifies the person holding them — all running locally.

From a phone photo to verified, structured data — on infrastructure you control.

TypeDocument AI · OCR · eKYC
LanguagesArabic · English
RunsLocally · on-premises
DeliverySDK · API · Desktop · Web · Mobile
Cisevo Document AI — Reads documents from imperfect photos and verifies the person holding them — all running locally.
AR + ENOCR models run in parallel
2liveness modes: active and passive
0external AI APIs required
5ways to deliver: SDK, API, desktop, web, mobile

Overview

An AI platform for document processing and eKYC. It corrects imperfect photos, reads Arabic and English with several OCR models, turns text into the fields each client needs, classifies Egyptian National IDs, and matches the document portrait to a live selfie with active and passive liveness — without depending on external AI services.

The problem

Real customers photograph documents at an angle, in poor light, on any phone — and onboarding still has to read every field correctly and confirm that the person is real and matches the document.

01

Reads photos that are far from perfect

Users do not need a scanner or a perfect shot. Before any text is read, the image is cleaned and straightened so documents photographed at an angle or in poor light can still be processed.

  • Noise reduction
  • Deskewing, perspective and geometric correction
  • Alignment
  • Lighting, brightness and contrast enhancement
Cisevo Document AI — document correction and OCR (illustration)
02

Several OCR models, side by side

Bilingual models read Arabic and English together, and Arabic-specialised models handle Arabic text. Teams can pick one model for a document type, or run several in parallel on the same image and compare their results.

03

From text to the fields you need

Recognising text and extracting data are separate steps. Post-processing services turn OCR output into the exact fields a client needs, following each document’s structure and business rules.

  • Egyptian National ID — front, back and the portrait
  • Passports
  • Invoices, shaped to the client’s workflow
  • Custom documents built on request
04

Egyptian National ID classification

A dedicated model trained on Egyptian National ID characteristics — such as the header “جمهورية مصر العربية” and how it is printed — classifies whether a card appears genuine or fake. It is specific to the Egyptian National ID and is not a check against any government database.

05

Face match and liveness

The portrait extracted from the ID is compared with a live selfie and a similarity score is returned. Active liveness asks the person to turn left, right or up; passive liveness works without any movement. Use either or both.

Cisevo Document AI — face match and liveness (illustration)
06

Runs locally, delivered your way

OCR, face recognition and liveness run on your own servers or devices, with no mandatory external AI service — suited to organisations with strict privacy or network rules. Delivered as an SDK, an API, or desktop, web and mobile applications.

Illustrations use a sample card design and a fictional, AI-generated person; the platform’s internal applications are not shown. Accuracy and performance figures will be published only after verified testing.

Next project

Cisevo Document Studio →

02

Contact

Open for new projects

Have an idea? Let’s build it.

Tell us what you want to make. We reply within one working day.