Auto enhance that decides
Grayscale, contrast, adaptive black & white, denoise or nothing at all — chosen from what the image actually looks like, and it tells you what it did.
Your documents don’t need to be uploaded to our server
Turn any image into editable text.
Take a photo, upload an image or scan a document. OCR2text extracts the text directly on your device.
No account. No AI. No document storage.
Five ways in
A phone, a laptop webcam, a file, a screenshot or a PDF — the same workspace handles all of them.
Capture, clean, recognise, correct, export
Most OCR sites give you an upload box and a spinner. The hard part is everything around the recognition.
- 1
Capture
Photo, webcam, file, paste or PDF page — several at a time if you like.
- 2
Clean
Straighten, flatten a photographed page, crop and enhance before recognition.
- 3
Recognise
Runs in a Web Worker so the page stays responsive. Cancel whenever you like.
- 4
Correct
Uncertain words are flagged and reviewed against the original crop.
- 5
Export
Copy, TXT, Markdown, DOCX, CSV, JSON or a searchable PDF.
Built for imperfect photographs
Because that is what real documents look like.
Flatten a photographed page
Drag four corners onto the page and it is warped back into a rectangle. Corners are detected for you where possible.
Review uncertain words
Step through the words the engine doubted, each with the original crop beside an editable field. Accept, edit or skip.
Extract just one thing
Drag a box around a serial number, an address or an error message and read only that. Fast, and often all you need.
Languages on demand
English by default. Add French, German, Irish, Hindi, Arabic and more, downloaded once and cached — with an honest note on how well each performs.
Tables, without pretending
Columns are inferred from real word positions. CSV export appears only when the structure holds up.
Works offline
Install it, and once the model is cached OCR2text keeps working with no connection at all.
One app, every device
A document scanner on a phone, a two-panel workspace on a laptop, and a proper layout on a tablet.
Nothing leaves the browser
No account, no upload, no scan history on a server, no AI API. Check the network panel yourself.
Common questions
Is my document uploaded anywhere?+
No. Recognition runs inside your browser using WebAssembly, so the image never leaves your device. The only thing downloaded from the network is the recognition model itself — a file of language data, not your page. You can watch this happen in your browser’s network panel.
Do I need an account?+
No account, no sign-in, no email. Close the tab and nothing about your document remains.
Does it work offline?+
Yes, once the app and your language model have been cached. Install OCR2text to your home screen or desktop and it keeps working on a plane or a train. The header says “Available offline” only when that is genuinely true.
How accurate is it?+
On clean printed text it is very good. On a phone photo of a curled page in bad light it is not magic — which is why OCR2text spends real effort on cleaning the image first, tells you which words it was unsure about, and lets you fix them against the original in a couple of taps.
Can it read handwriting?+
Not reliably. The engine is built for printed text. You can try, and sometimes neat handwriting comes through, but expect to do a lot of correcting.
What about tables?+
OCR2text reconstructs columns from where words actually sit on the page, and offers a CSV export when the layout is consistent enough to defend. When it is not, it says so instead of inventing a tidy-looking grid.
What does it cost?+
Nothing. Recognition happens on your device, so there is no per-page server cost to pass on.
Extract text from something
No account, no upload, no waiting for a server. Open the workspace and drop a page in.