About
Hebrew documents reach Excel broken. We built a tool that gets them there whole.
OCR-AI is an independent product built in Israel, for the documents people actually receive: invoices, receipts, delivery notes and scanned pages, in Hebrew, in English and in both at once.
01Why
The problem it exists for
Anyone who has converted a Hebrew invoice to text knows the result: words in reverse order, numbers split in the middle of a line, and a CSV that opens in Excel as rows of symbols. Many OCR tools were built for English, with Hebrew added later.
We wanted the opposite: drop a document and get data you can use straight away. Text in the right reading order, fields such as supplier, date, VAT and allocation number, tables in the right columns, and a sheet that opens right-to-left.
And we wanted it to be easy to try: no signup, no card, 5 free pages every day. People who need more buy a page pack; people who don't, don't pay.
02Principles
What we hold ourselves to
- 01
The product is the page
There is no signup form between a visitor and their document. Drop it, see the result, then decide.
- 02
Hebrew first
Reading order, Israeli document fields, right-to-left Excel and a CSV that opens in Hebrew. Not a translated English product.
- 03
Your documents are not our data
The file, the image and the text are not stored on the server. Your history stays in your browser.
- 04
No numbers we did not measure
We will not publish accuracy figures we did not measure. Values the engine was unsure of are marked for review, and we say plainly where to check.
- 05
Pay for what you use
A one-time page pack, no subscription. A failed read is not counted.
03What it is not
So there are no surprises
- It is not accounting software. The result is a file you check and import into your own system.
- It is not an archive. Documents are not stored with us, and the history in your browser is a convenience, not a backup.
- There is no API or direct integration yet. For large volumes, write to us.
- Automatic reading can be wrong, especially with handwriting and weak photos. That is why every value can be checked against the page and corrected.
The best way to get to know it
Open a document. It takes less time than reading this page.