Question 1

What happens to my documents?

Accepted Answer

They stay in your workspace so you can re-open, export and chat with them, and they remain there until you delete them. Processing runs on isolated GPU workers, the uploaded files sit in private storage with no public address at all — your browser opens them through short-lived links issued to your account, and row-level security keeps every other account out — and nothing you upload is used to train models.

Question 2

How accurate is the extraction?

Accepted Answer

It depends on the page, and we deliberately do not publish a single headline percentage — one number covering clean PDFs, creased receipts and handwriting would tell you nothing. What we can tell you is what happens when a page reads badly: a low yield automatically triggers a second engine, and a refinement that drifts too far from the source text is thrown away in favour of the raw OCR output.

Question 3

What happens if I use up my monthly pages?

Accepted Answer

Processing pauses until your period rolls over or you upgrade — there is no surprise overage bill. Free accounts get 10 pages a month and Pro gets 250. A document that does not fit into your remaining pages is rejected whole rather than processed halfway, and a run that fails is never charged at all.

Question 4

Which languages are supported?

Accepted Answer

Eleven: English, Turkish, Spanish, French, German, Italian, Portuguese, Russian, Chinese, Japanese and Korean. You choose the language when you upload — nothing is auto-detected — and both OCR and the refinement step commit to that choice, which is what keeps a Turkish invoice from being read as though it were English.

Question 5

What is chat with documents?

Accepted Answer

Your text is split into roughly 1,000-character chunks, embedded and stored in pgvector. When you ask a question, the closest chunks are retrieved and the answer is written from those, along with the file they came from. Ask a 50-page contract what its termination terms are instead of reading it. Document-scoped chat is a Pro feature.

Question 6

How is this different from ordinary OCR?

Accepted Answer

Ordinary OCR hands you characters. Here the document is first routed to the engine that suits it — a digital PDF is read from its text layer and never touches a GPU — then refined by a model that is not allowed to see emails, URLs or banking identifiers, and finally indexed so the document can answer questions.

Question 7

Can it read handwriting?

Accepted Answer

Yes. Handwritten pages go to Google Vision first, and a page it cannot read is handed to OpenAI Vision rather than given up on. If a document you did not flag as handwritten comes back suspiciously empty, handwriting mode is retried automatically before anything is reported as failed.

Question 8

Which file formats can I upload?

Accepted Answer

PDF, JPG, PNG, WebP and TIFF. Scans and phone photos are deskewed before OCR, so a page photographed at an angle still reads correctly.

Question 9

Can I get an invoice for my subscription?

Accepted Answer

Yes. Payments run through Polar as merchant of record: your billing details are collected at checkout and Polar issues the invoice for each payment. Card details never touch this application.

Question 10

Can I cancel at any time?

Accepted Answer

Yes, in one click and with no commitment. Cancelling stops the renewal rather than cutting you off — Pro stays active until the end of the period you have already paid for.

Question 11

Do you offer refunds?

Accepted Answer

A period that has already started is not refunded — cancelling ends the renewal and you keep Pro until that period runs out. Statutory refund rights in your own jurisdiction still apply and override this. If something went wrong on our side rather than yours, write to support@ocrcon.com and we will sort it out.

Frequently Asked Questions