How to Turn a Photo of a Document Into Editable Text
You take a photo of a contract clause, a receipt, a whiteboard full of meeting notes — and then try to select the text to copy it somewhere, search it, or translate it. Nothing highlights. Long-press, drag-select, none of it works, because what you’re looking at is a picture, not text.
The next move is either retyping the whole thing by hand or finding a tool that can “read” it for you. The second option is obviously less painful — this walks through how it actually works, how accurate it is, and whether your photo ends up anywhere it shouldn’t.
Why you can’t just select text in a photo
To a computer, a photo is just a grid of pixels. It has no idea whether those pixels happen to form letters a human would recognize — there’s no underlying “this says X” data attached to the image, the way there is in a Word document or a PDF with a real text layer. Selecting and copying text requires that the content actually exist as characters, not as pixels.
Turning pixels into characters is what OCR (optical character recognition) does — it’s the same basic task as a human reading a photo, just automated.
How accurate is it, really
OCR isn’t 100% accurate, and how well it works depends heavily on the photo itself.
- Works well: printed text, even lighting, sharp focus, minimal tilt — recognition accuracy is quite high here, for both English and Chinese.
- More error-prone: handwriting, blurry or dimly lit or tilted photos, dense tables — misread characters go up noticeably, and formatting like underlines or boxed fields often doesn’t survive the trip.
So treat OCR output as a starting point that saves you from retyping everything, not as a finished, ready-to-use result. If the photo contains amounts, dates, contract terms, or ID numbers, check the recognized text against the original photo before you actually use it — don’t just copy-paste it straight into something that matters.
How to extract the text
- Open Extract Text (OCR) and drop in the photo — or several at once, no need to process them one by one.
- Click “Start recognition.” Everything runs locally in your own browser — the photo is never uploaded to any server. That matters more than it sounds when the photo is a contract, a receipt, or an ID: you’re not handing that content to some third-party server you know nothing about.
- Each photo comes back as a block of text you can copy directly, or download as a .txt file — batches can be downloaded as a single zip.
If what you actually have is a PDF rather than a photo (a scanned document, or one where the text was flattened into shapes during export and has no real text layer), batch PDF to Text is the better starting point — it detects the missing text layer on its own and automatically falls back to the same OCR engine, no need to convert to an image yourself first.
Before you use the result
The tool shows a confirmation prompt before every download, asking you to double-check the content first — that’s not there for show. OCR is recognition, not extraction, and it isn’t infallible. Give the result a quick check against the original photo, especially anything involving amounts, dates, or contract terms, before putting it to real use.