Image to Text (OCR)
Extract text from a photo, screenshot, or scan — free, unlimited, and never uploaded.
By The Paper Room Editorial Team — Image Tools
Image to Text (OCR) workspace
Workflow guide
Add files, review the accepted files and settings, then process and download a downloadable file.
Your files and inputs stay in this browser unless you choose an account-backed feature.
How to use Image to Text (OCR)
Upload a photo, screenshot, or scanned page and pull the text out of it. Recognition runs entirely in your browser using a WebAssembly build of Tesseract, the long-established open-source OCR engine, so the image is never uploaded to a server and there is no account, no page limit, and no queue.
That local-first design is the reason this can be free and unlimited where hosted OCR services meter by the page. It has one honest cost: the first time you read a given language, the browser downloads that language's trained data file, which is tens of megabytes. After that it is cached, and later images start immediately. A dense page takes a few seconds to process rather than being instant.
Accuracy depends far more on the image than on the settings. Sharp, straight, high-contrast text on a plain background reads close to perfectly; a blurry photo taken at an angle in poor light will not, no matter which tool you use. If the result comes back empty or garbled, re-shoot or re-scan before assuming the text is unreadable — and check that the selected language matches what is actually on the page, since a Latin-alphabet model cannot read Cyrillic or Devanagari.
By The Paper Room Editorial Team — Image Tools
Frequently asked questions
Is my image uploaded anywhere?▼
No. The recognition engine runs as WebAssembly inside your own browser, so the image stays on your device for the entire process. You can confirm this by disconnecting from the internet after the page and language pack have loaded — the tool keeps working.
Why is the first scan slow?▼
The first read in any language downloads that language's trained data file, which is tens of megabytes. Your browser caches it, so every later scan in the same language skips that step and starts straight away.
Which languages are supported?▼
English, Spanish, French, German, Portuguese, Italian, Dutch, Russian, Arabic, Hindi, Simplified Chinese, Japanese, and Korean. Pick the language that matches the text in your image — a model trained on one script cannot read another.
Can it read handwriting?▼
Generally not well. Tesseract is trained on printed text and does poorly on cursive or casual handwriting. Neat block capitals sometimes work. For handwritten notes, expect to correct the output.
Why did it return nothing or gibberish?▼
Almost always an image-quality issue: blur, low contrast, a steep angle, or a skewed scan. Re-take the photo straight-on with even lighting, or rescan at a higher resolution. Also confirm the language selector matches the text — the wrong language produces plausible-looking nonsense rather than an error.
Does it keep the original layout?▼
Partly. Line breaks and paragraph order are usually preserved, but multi-column layouts, tables, and text wrapped around images often come out in reading order rather than visual order. Complex layouts are worth a quick check against the original.