Key takeaways
- Every tool here reads printed Arabic; handwriting is where they differ. Amazon Textract recognises handwriting in English only, Arabic is not in the handwriting list of Azure’s Read model, and ABBYY FineReader Engine does not support Arabic handwriting except handprinted digits.
- Three questions settle most projects in the UAE: where the scans may be processed (a public cloud region, a UAE region or your own servers), how mixed Arabic and English pages come out, and who checks the words the engine is unsure of.
- Cloud OCR is priced per page: $1.50 per 1,000 pages at Amazon Textract (Detect Document Text) and Google Cloud Vision after their free allowances. ABBYY FineReader PDF is desktop software from $99 a year, and Tesseract is free and open source.
- To keep scans on your own servers, Tesseract and ABBYY run where you install them, Azure offers its Read OCR as a container, Google offers OCR in Google Distributed Cloud, and Wathiqa runs on your servers or in a UAE cloud region.
How to choose an Arabic OCR tool
- Printed or handwritten: printed Arabic is widely supported, handwriting much less, so test with your own forms and letters.
- Mixed pages: Arabic and English on the same page, tables, stamps and signatures, and numbers in Arabic-Indic or Western digits.
- Where the scans are processed: a public cloud region, a UAE cloud region or your own servers. Government records and personal data often need one of the last two.
- Confidence and review: a confidence score for every word, and a screen where a person corrects the uncertain ones before the text is used.
- Output: searchable PDFs, plain text, or fields such as names, dates and amounts for a workflow.
- Volume and pricing: per page for cloud APIs, per user or per licence for software.
- Where the text goes next: a one-off conversion, or a document management system that keeps the scan and the text together with search, retention and an audit log.
The six tools compared
“Not stated” means we did not find it on the vendor’s own website or documentation in November 2026, so ask the vendor. Prices are as published, and they change.
| Tool | Where it runs | Pricing | Printed Arabic | Arabic handwriting |
|---|---|---|---|---|
| Wathiqa (Birbal AI) | Your servers or a UAE cloud region | Arabic OCR in the Business plan, AED 110 per user a month, billed annually; on-premise quoted | Yes, Arabic and English, with a confidence score for every word and a review screen | Depends on the handwriting; measured on your own samples in a demo |
| ABBYY FineReader | FineReader PDF on your Windows or Mac computer; FineReader Engine inside your own software | FineReader PDF for Windows: Standard $99 a year, Corporate $165 a year | Yes, among 201 OCR languages | Not supported in FineReader Engine, except handprinted digits |
| Google Cloud (Vision API, Document AI) | Google Cloud; OCR also in Google Distributed Cloud for your own data centre | Vision: first 1,000 units a month free, then $1.50 per 1,000 | Yes | Document AI reads handwriting in 50 languages; Arabic not named |
| Microsoft Azure Document Intelligence | Azure; Read OCR also as a container on your servers | 500 pages a month free, then billed per 1,000 pages | Yes, in the Read model | No: Arabic is not in the handwriting list |
| Amazon Textract | AWS cloud | Detect Document Text $0.0015 a page for the first million pages a month | Yes, among 40+ languages | No: handwriting in English only |
| Tesseract | Wherever you install it | Free and open source (Apache 2.0) | Yes, with its Arabic language data | Not stated; built for printed text |
The tools, one by one
Wathiqa (Birbal AI)
Wathiqa is Birbal AI’s document management system, with Arabic and English OCR built in. Scans, emails and PDFs are read as they come in, every page and word gets a confidence score, and a review screen shows the scan beside the extracted text so your team can correct any word the AI is unsure of before it is used. The text is then searchable by meaning, kept under retention rules and recorded in a tamper-evident audit log. It runs on your own servers, including isolated networks, or in a UAE cloud region. Arabic OCR is in the Business plan at AED 110 per user per month, billed annually, and on-premise licences are quoted. See Wathiqa.
ABBYY FineReader
ABBYY FineReader PDF is desktop software that turns scans into searchable PDFs and editable files. Version 15 lists 201 OCR languages, including Arabic, and ABBYY states that version 16 improved Arabic recognition. FineReader PDF for Windows costs $99 a year (Standard) or $165 a year (Corporate, which adds automated conversion of up to 5,000 pages a month). For developers, FineReader Engine adds OCR to your own software; its language table lists Arabic with a full dictionary, but no Arabic handwriting recognition except handprinted digits. A good fit for a team that converts documents on its own computers.
Google Cloud
Google’s Cloud Vision API reads Arabic in images and PDFs, and Document AI’s Enterprise Document OCR covers more than 200 languages, with handwriting in 50; Google lists Arabic for its Form Parser, Layout Parser and Custom Extractor. Vision’s text detection is free for the first 1,000 units a month, then $1.50 per 1,000 units up to five million. For organisations that cannot use the public cloud, Google lists Arabic among the OCR languages of Google Distributed Cloud, its edition that runs in your own data centre with no connection to the internet.
Microsoft Azure Document Intelligence
The Read model in Azure Document Intelligence lists Arabic among its printed-text languages; its handwriting list covers English, Chinese (Simplified), French, German, Italian, Japanese, Korean, Portuguese and Spanish, not Arabic. The Read OCR is also available as a container, so it can run on your own servers. Billing is per page: 500 pages a month are free, then you pay per 1,000 pages, and commitment tiers include a disconnected container for high volumes.
Amazon Textract
Amazon Textract detects printed text, including right-to-left and mixed-script documents, in more than 40 languages, Arabic among them, and returns a confidence score for every word and line. It recognises handwriting in English only, and its forms, queries, invoice and ID features are limited to English or a few European languages. Detect Document Text costs $0.0015 a page for the first million pages a month in US West (Oregon), and new AWS customers get 1,000 pages a month free for three months. A good fit for printed Arabic in a workflow already built on AWS.
Tesseract
Tesseract is a free, open-source OCR engine (Apache 2.0 licence) that extracts printed text from images. It has Arabic language data and an Arabic script model, runs wherever you install it and has no licence fee. It has no review screen, document storage or support contract, so your team builds and maintains everything around it. A good fit for developers working with clean, printed scans.
What Arabic OCR costs
Cloud OCR is priced per page or image. After their free allowances, Amazon Textract’s Detect Document Text and Google Cloud Vision’s text detection both cost $1.50 per 1,000 pages, falling at higher volumes, and Azure Document Intelligence gives 500 free pages a month before billing per 1,000 pages. Desktop software is priced per user: ABBYY FineReader PDF for Windows costs $99 or $165 a year. Tesseract is free, but you pay in engineering time.
These prices cover reading the page. They leave out what happens next: a person checking uncertain words, keeping the text with the original scan, making it searchable and keeping it for the legal retention period. Our guide to document management systems in the UAE compares the systems that do that.
Where Birbal AI fits
Birbal AI makes Wathiqa for organisations in the UAE that need Arabic and English scans read, checked by a person where the engine is unsure, and kept as searchable records on their own servers or in a UAE cloud region. If you only need to convert a few printed documents on one computer, ABBYY FineReader PDF is simpler, and if you are building your own pipeline on AWS, Google Cloud or Azure, their OCR services are priced per page.
Questions and answers
What is the best Arabic OCR?
It depends on what you scan and where it may be processed. All six tools here read printed Arabic. For handwriting, Amazon Textract and Azure’s Read model do not list Arabic, and ABBYY FineReader Engine supports only handprinted digits, so test with your own samples. For records that must stay in the UAE, choose a tool that runs on your servers or in a UAE region, such as Wathiqa.
Can OCR read handwritten Arabic?
Support is limited. Amazon Textract recognises English handwriting only, Arabic is not in Azure Document Intelligence’s handwriting list, and ABBYY FineReader Engine does not support Arabic handwriting except handprinted digits. Google’s Document AI reads handwriting in 50 languages, but we did not find Arabic named among them. Test with your own forms before you choose.
Is there free Arabic OCR?
Tesseract is free and open source and reads printed Arabic with its Arabic language data. Cloud services also have free allowances: the first 1,000 units a month on Google Cloud Vision, 500 pages a month on Azure Document Intelligence, and 1,000 pages a month for three months on Amazon Textract for new AWS customers. Our free Arabic OCR tool runs Tesseract in your browser, so nothing is uploaded.
Is there an Arabic OCR model made in the UAE?
Yes. On 7 October 2026 the Technology Innovation Institute (TII) in Abu Dhabi launched Falcon-OCR-Arabic, a lightweight model that extracts Arabic text and structured information, including tables and document sections, from images and documents. TII’s Falcon Arabic OCR site reads print, layout and handwriting, including multi-page PDFs. As with any new model, test it on your own documents first.
Is it safe to use a free online Arabic OCR website?
For public text, usually. A free website uploads your file to its own servers, so for contracts, IDs, invoices or other personal data, check where the site keeps files and when it deletes them, or use a tool that runs on your own device, such as our free Arabic OCR, or on your own servers.
Which Arabic OCR tools can run on our own servers?
Tesseract and ABBYY FineReader run where you install them, Azure offers its Read OCR as a container, Google lists Arabic OCR in Google Distributed Cloud, and Wathiqa runs on your servers, including isolated networks, or in a UAE cloud region. Amazon Textract is a cloud service.
How accurate is Arabic OCR?
It depends on scan quality, fonts, diacritics, stamps and handwriting, so the only reliable figure is one measured on your own documents. Amazon Textract and Wathiqa return a confidence score for every word, which lets a person check the uncertain ones.
What is Arabic OCR used for?
Making scanned contracts, invoices, letters and archives searchable, pulling names, dates and amounts into workflows, and turning paper records into a digital archive. The text is only as useful as the place it is kept, so most organisations pair OCR with a document management system.
Sources
- Amazon Textract FAQs (AWS)
- Amazon Textract pricing (AWS)
- Supported languages in Amazon Textract (AWS)
- OCR language support in Document Intelligence (Microsoft)
- OCR overview, Azure AI Vision (Microsoft)
- Azure Document Intelligence pricing (Microsoft)
- FineReader PDF pricing (ABBYY)
- FineReader PDF 15 supported languages (ABBYY)
- FineReader PDF 16 upgrade (ABBYY)
- FineReader Engine predefined languages (ABBYY)
- Cloud Vision supported languages (Google Cloud)
- Cloud Vision pricing (Google Cloud)
- Detect text in files (PDF/TIFF), Cloud Vision (Google Cloud)
- OCR with Google Cloud (Google Cloud)
- Document AI languages (Google Cloud)
- OCR languages in Google Distributed Cloud (Google Cloud)
- Installing Tesseract (Tesseract documentation)
- Tesseract data files (Tesseract documentation)
- TII launches Falcon-Emirati alongside new Arabic AI models, 7 October 2026 (WAM)
- Falcon Arabic OCR (Technology Innovation Institute)
Birbal AI is an independent company based in Abu Dhabi. Facts about third parties come from the sources above; check them for the latest position.