Hebrew OCR tools compared: which one fits your documents
We make OCR-AI. Facts about other tools come from their own pages, checked on 6 November 2026. Each row links to its sources.
In short
- A page or two of printed Hebrew: i2OCR, FastOCR, DocuPipe and OCR-AI are free within their limits, and Google Docs works with any Google account for files up to 2 MB.
- A scanned PDF to edit in Word: check that the tool runs OCR and names Hebrew, because some converters only read PDFs that already contain text.
- Tables and invoices to Excel: OCR-AI, Kovetz Pro and ABBYY FineReader PDF say they pick up tables, and all three export to Excel. For handwriting, Kolmus (Torah novellae) and KtavAI are built for it.
01Side by side
What each tool's own pages say
Each fact is what the vendor's own page said on 6 November 2026, with prices in the currency the page showed. "Not stated" means the page did not say. "Hebrew named" means the vendor's page names Hebrew as supported. We did not run files through the other tools.
| Tool | Free | Paid | Scanned PDF | Output | Tables | Your file |
|---|---|---|---|---|---|---|
| OCR-AIWeb, plus an API and an MCP server for AI agentsHebrew namedSources: 1 2 3 4 5 6 | 5 pages a day per device, up to 5 pages from each document, no signup. There is also a limit per network | One-time packs: $9 for 100 pages, $29 for 500, $89 for 2,000 (₪29, ₪99, ₪299 on the Hebrew site). Pages stay valid 365 days from the latest purchase | Yes. Each PDF page is turned into an image in your browser and read on its own | Text, Word (.docx), Excel, CSV, JSON | Word tables with a header row. In Excel, a sheet per table; a table that continues on the next page with the same columns joins one sheet | The file, the page image and the text are not stored on our servers; results stay in your browser. The cloud provider that runs the reading engine may keep requests for a limited time to detect abuse |
| Google Drive / Google DocsWeb: upload to Google Drive on a computer, then Open with > Google DocsHebrew namedSources: 1 2 3 | With any Google account. Google says the file should be 2 MB or smaller | No separate price stated | Yes: PDFs, multipage too, and JPEG, PNG or GIF photos | A Google Doc, which you can download in other file types | Google says lists, tables, columns and footnotes are "not likely to be detected" | Works on a file that sits in your Google Drive |
| Adobe Acrobat onlineFree web tools; Acrobat Standard and Pro are paid plansHebrew not namedSources: 1 2 3 4 5 6 | PDF to Word: free with a daily limit (the number is not stated); after the first use you may be asked to sign in to download. Online OCR: sign in to download the result | Acrobat Standard ₪88 a month, Pro ₪105 a month (monthly plan, Adobe's Israeli page) | Adobe says Acrobat's OCR reads scanned PDFs. Its online OCR page, and the language list it links to, do not name Hebrew | Word (.docx or .doc) or RTF; the online OCR returns a searchable PDF. On Adobe's forum, its community team called Hebrew turning left to right in exports to Word a "known limitation" (28 April 2026) | Adobe says tables generally carry over to Word, and complex layouts may need cleanup | Handled on Adobe's servers and deleted unless you sign in to save it |
| ABBYY FineReader PDFDesktop app for Windows and MacHebrew namedSources: 1 2 3 4 | 7-day trial, no payment required | Subscription: Mac $69 a year; Windows Standard $99 a year or $16 a month; Corporate $165 a year | Yes | Word, Excel, PowerPoint, PDF and PDF/A, TXT, CSV, HTML, RTF, ODT, EPUB (Windows list) | ABBYY says it extracts text and tables from PDFs | Not stated (an installed app) |
| i2OCRWebHebrew namedSources: 1 2 3 | Free, no registration; one image or one PDF page per run. Images up to 20 MB, PDFs up to 200 MB | Bulk OCR is a paid service; price on request | Yes, one page at a time | Text, Word (DOC or DOCX), HTML, searchable PDF | i2OCR says it does not guarantee tables or the original formatting | Uploads and extracted text are deleted automatically within 30 minutes |
| i2PDF Hebrew PDF to WordWebHebrew namedSources: 1 2 | Free, no account; one PDF up to 100 pages and 100 MB | None: i2PDF says all its tools are free | No. The tool does not run OCR, so it needs a PDF that already has text | Word (.docx). i2PDF says Hebrew can come out as right-to-left paragraphs | Supported tables are rebuilt as Word tables | Deleted from its servers within 30 minutes after processing |
| Convertio OCRWebHebrew namedSources: 1 2 3 | 10 pages without an account; for more pages it asks you to sign up | Prepaid OCR packs, from $7.99 for 100 pages | Yes: scanned documents and images, PDF included | Word, Excel, searchable PDF, TXT, RTF, CSV, PowerPoint, EPUB and more | Not stated (Excel and CSV are among the outputs) | Input files are deleted right after conversion, results after 24 hours |
| Kovetz (קובץ)Web (Israeli)Hebrew namedSources: 1 2 3 | 3 documents a day, no signup; files up to 25 MB and 100 pages. From a scan, only the first page converts free, as a sample | Pro: ₪19 a month or ₪144 a year | With Pro: scanned PDF to Word or Excel, up to 30 pages per conversion | Word, Excel, PowerPoint, HTML, text, searchable PDF | Detected tables become Word tables; tables from scans go to Excel with column detection (Pro) | Deleted from its servers within 10 minutes |
| DocuPipe Hebrew OCRWeb tool, plus an APIHebrew namedSources: 1 2 | No signup: the first 3 pages per run, files up to 200 MB, a daily quota per network. A free account gets 100 credits a month | Business: $99 a month for 2,500 credits; reading a page costs 1 credit | Yes: scans, phone photos and PDFs | Searchable PDF, text and JSON | A separate free PDF-to-Excel tool puts each table on its own sheet | Encrypted in transit and at rest, never used to train models; how long files are kept is not stated |
| FastOCRWeb, plus free Mac and Windows apps and an APIHebrew namedSources: 1 2 3 | Image OCR is free (try it without an account); PDFs need a free account, 20 pages a month | Page packs from $4.99 for 500 pages (one-time), or Pro at $9.99 a month for 1,500 pages | Yes, with a free account | Word (.docx), text, searchable PDF | FastOCR says table grids are not recreated in the Word file | Results stay available for 30 days, then are deleted automatically; the desktop apps work offline |
| TesseractOpen-source OCR engine and command-line program; no graphical appHebrew namedSources: 1 2 3 | Free and open source (Apache 2.0); needs installing | None | Reads images. It does not open PDF files itself, so a PDF is turned into images first | Text, hOCR (HTML), PDF, PDF with only an invisible text layer, TSV, ALTO, PAGE | Not stated | Runs on the computer where you install it |
02By job
Which tools fit which job
A page or two, quickly and free
i2OCR (one page per run, no registration), FastOCR (images, no account needed to try), DocuPipe (the first 3 pages, no signup), Convertio (10 pages without an account) and OCR-AI (5 pages a day, no signup). Google Docs also reads Hebrew, with any Google account, for files up to 2 MB.
A scanned PDF to Word
Tools whose pages name Hebrew and offer OCR to Word: OCR-AI, Convertio, i2OCR (a page at a time), FastOCR (free account), Kovetz Pro (up to 30 pages per conversion) and ABBYY FineReader PDF. Google Docs gives a Google Doc you can download. i2PDF has no OCR, so it suits PDFs that already have text.
Tables and invoices to Excel
OCR-AI (a sheet per table; for invoices, the fields and the line items too), Kovetz Pro (tables from scans, with column detection), ABBYY FineReader PDF (exports XLSX) and DocuPipe's separate PDF-to-Excel tool.
Books, niqqud and older print
KtavAI (ktavyad.com) says it supports Rashi script and multi-column print. FastOCR says it reads niqqud when the page has it, and i2OCR says clear niqqud may be recognized. Check the vowel marks yourself.
Handwriting
Kolmus (hakulmus.co.il) is built for handwritten Torah novellae: you proofread line by line, then export to Word. KtavAI (ktavyad.com) transcribes Hebrew handwriting to TXT or DOCX. FastOCR says it does not read cursive Hebrew yet. With any tool, expect to proofread.
Developers and AI agents
Tesseract (open source, with a Hebrew model), IronOCR (a paid .NET library with a Hebrew language pack), the FastOCR and DocuPipe APIs, and the OCR-AI API and MCP server (a new account gets 100 free pages once, with its first key).
03Before you choose
Five checks to run on your own file
- 01A line that mixes Hebrew with an invoice number, a date and an English word: does each part stay whole and in its place?
- 02Final letters (ך ם ן ף ץ): do they stay final, or turn into the regular forms?
- 03A table: does it come out as a table, with the columns in the right order, or as loose lines of text?
- 04Excel: does the sheet open right to left, with the first column on the right?
- 05CSV: does it open in Excel as Hebrew, not as a row of symbols?
04In fairness
When OCR-AI is not the pick
Cursive handwriting
OCR-AI is built for print, and handwriting is read partly at best. Kolmus (for Torah novellae) and KtavAI are built for handwriting.
Text dense with niqqud
Printed, legible niqqud is kept, but small or faded vowel marks read less reliably. If every vowel matters, check each line.
Rashi script and older print
Our pages make no promise for them. KtavAI says it supports Rashi script.
A copy that keeps the look
Our Word file is for editing the content: fonts, colours, images, newspaper-style columns and exact positions on the page are not kept.
05Method
How we compiled this
We read each tool's own pages on 6 November 2026: product pages, price lists, help pages and FAQs. For Adobe we also used a reply from Adobe's community team on Adobe's own forum.
Lists by other companies only told us where to look. Where a vendor's page did not say something, the table says "not stated". We left out tools whose own pages do not name Hebrew for OCR, except Adobe Acrobat: its row stays, and it shows that Adobe's online OCR pages do not name Hebrew.
We did not run files through the other tools, and this page gives no scores, ratings or accuracy figures.
Prices and limits change, so check the vendor's page before you pay. Found a mistake or a change? Write to us and we will correct it.
Questions
Common questions
What is the best free Hebrew OCR?
There is no single answer, and we don't rank tools. The free options differ by limit: i2OCR reads one image or page per run, FastOCR images without an account, DocuPipe the first 3 pages per run, Convertio 10 pages without an account, and OCR-AI 5 pages a day without signup. Google Docs works with any Google account, for files up to 2 MB. Tesseract is free and open source, but you install it. Run the five checks above on your own file, and keep the tool whose result needs the fewest fixes.
Can Google Docs read Hebrew?
Yes. Hebrew is on the list of languages Google Drive detects. Upload the PDF or image to Google Drive on a computer, right-click it and choose Open with > Google Docs. Google says the file should be 2 MB or smaller, and that tables, lists and columns are not likely to be detected.
Why does Hebrew come out backwards after converting a PDF to Word?
A PDF sometimes stores letters in the order they are drawn on the page, not the order they are read. A converter that copies that order reverses Hebrew lines, and numbers or English words inside a line swap places. On Adobe's forum, Adobe's community team called Hebrew coming out left to right in Acrobat's export to Word a known limitation (28 April 2026). OCR-AI reads each page the way it looks and writes the text in reading order. If you already have reversed text, the reversed-Hebrew fixer on this site runs in your browser and keeps numbers and English whole, and Kovetz has a Fix Hebrew tool that runs in the browser too.
Which Hebrew OCR keeps tables and exports to Excel?
From the vendors' own pages: OCR-AI exports Excel with a sheet per table and puts tables into Word as Word tables. Kovetz Pro turns tables from scans into Excel with column detection. ABBYY FineReader PDF exports XLSX and says it extracts tables. DocuPipe has a separate free PDF-to-Excel tool, a sheet per table. Convertio lists Excel and CSV as outputs but does not say how it handles tables. Google says tables are not likely to be detected, and i2OCR does not guarantee them.
Does any OCR read Hebrew handwriting?
Some tools say they do. Kolmus is built for handwritten Torah novellae, and KtavAI (ktavyad.com) transcribes Hebrew handwriting. i2OCR says handwritten Hebrew is supported but less reliable than print. FastOCR says it does not read cursive Hebrew yet, and OCR-AI is built for print and reads handwriting partly at best. Whatever you use, proofread against the page.
Is there a Hebrew OCR API?
Yes, several. Tesseract is open source and has a Hebrew model. IronOCR is a paid .NET library with a Hebrew language pack. FastOCR and DocuPipe offer APIs. OCR-AI has an API and an MCP server for AI agents: a new account gets 100 free pages once, with its first key, and then uses the same page packs as the site.
Is it safe to upload documents to an online OCR?
Read what each tool says it does with your file; the table has a column for it. Kovetz says files are deleted within 10 minutes, i2OCR and i2PDF within 30 minutes, and Convertio deletes results after 24 hours. FastOCR keeps results for 30 days, and Google works on a file in your Drive. OCR-AI does not store the file, the page image or the text on its servers, and results stay in your browser; the cloud provider that runs the reading engine may keep requests for a limited time to detect abuse. If a file must not leave your computer, FastOCR says its desktop apps work offline, and Tesseract runs where you install it.