About FastOCR

FastOCR was built out of frustration.

I am a software engineer who also spends a lot of time researching history. That means reading old books — scanned PDFs of texts in Urdu, Arabic, Farsi, and other languages that most software barely supports. I needed to search through these documents, copy passages, and organize my notes. Simple stuff.

But every OCR tool I tried either did not work or made it painful:

• Free online tools cap you at 2-3 pages. A scanned book has hundreds.
• Desktop apps require downloading heavy software, and most still produce garbage output for non-English text.
• The few tools that handle Arabic or Urdu script cost $20+/month and still get the text direction wrong in the output PDF.
• None of them give you a proper searchable PDF — they just dump raw text with broken formatting.

I kept thinking: Google Vision AI can read these scripts with 95%+ accuracy. Why is there no simple tool that lets you upload a PDF, runs Vision AI on it, and gives you back a searchable version? No downloads, no complicated setup, no 3-page limit.

So I built one.

What FastOCR does differently

FastOCR is a cloud-based OCR tool powered by Google Vision AI. You upload an image or PDF, and it extracts the text with high accuracy — including right-to-left languages like Urdu, Arabic, and Farsi that most tools struggle with.

For PDFs, it does something most free tools cannot: it creates a searchable PDF. The original scan stays intact, but an invisible text layer is added on top. You can press Ctrl+F and find any word. You can select and copy text. The text direction is correct, even for RTL scripts.

There is no page limit on what the tool can process. A 30-page document takes about a minute. A 500-page book works too — it just takes longer.

The numbers

How it works under the hood

FastOCR runs entirely on AWS serverless infrastructure. When you upload a file, it goes to a secure S3 bucket. A Lambda function sends each page to Google Vision AI for text recognition. The results are assembled into a searchable PDF using PyMuPDF and ReportLab, with special handling for right-to-left text positioning. The whole pipeline is automated — no human ever sees your documents.

Your files are processed and stored temporarily for download. They are not used for training, advertising, or anything else.

Why it is free

Image OCR is free and always will be — no registration, no limits. PDF processing requires a free account (3 PDFs per month). If you need more, there is a Pro plan at $9.99/month for 100 PDFs.

I built this because I needed it. If it helps other researchers, students, and professionals who work with non-English documents, even better.