Image to Text (OCR)

Turn images into editable text without uploading them anywhere. Drop in a screenshot, a photo of a document, a scan or a sign, and a local OCR engine reads the text right on your device — in 20 languages, including mixed-language pages. Edit the result, copy it to the clipboard, or download it as a .txt file. The OCR engine downloads once on first use and is cached; your images never leave your browser.

Privacy-first: every image is processed locally in your browser. Nothing is uploaded, stored or transmitted to any server.

PixelVault turns images into text with OCR that runs entirely in your browser. Screenshots of error messages, photos of book pages, scanned letters, receipts, signs, whiteboards — drop them in and the text comes out the other side, selectable and editable. Unlike the usual “image to text” websites, nothing is uploaded: the recognition engine itself runs on your device.

Twenty recognition languages are available, from English, Spanish and German to Arabic, Hindi, Japanese, Korean and both Simplified and Traditional Chinese. You can enable up to three at once for mixed-language documents. Each language pack is a small one-time download (about 1–3 MB) that your browser keeps cached.

The result appears in an editable text box together with a confidence score, so you can fix the odd character before using it. Copy everything to the clipboard with one click, or download it as a .txt file named after your image.

How it works

PixelVault uses Tesseract, the most widely deployed open-source OCR engine, compiled to WebAssembly so it can run inside your browser. On first use it downloads the engine (about 5 MB) and the trained data for your selected languages — you are told the exact size before anything is fetched. The image is analysed in a background thread: the engine finds lines and words, recognizes characters against its language model, and returns plain text plus a confidence estimate. Everything is cached locally, so from the second run onwards it also works offline.

Why local processing matters

The images people run through OCR are usually the sensitive ones: contracts, invoices, ID documents, medical letters, private chats. Uploading them to an OCR service means those documents sit, at least briefly, on someone else’s infrastructure. Local OCR removes that exposure completely — the pixels never leave your machine, and there is no service on the other end to log, cache or train on your documents.

Supported formats

  • PNG / JPG — Screenshots and photos — the most common OCR sources.
  • WebP / AVIF / BMP / GIF — All accepted as input.
  • HEIC — iPhone photos of documents are decoded locally first.
  • TXT — The output: plain text, ready to copy or download.

Common use cases

  • Copy an error message out of a screenshot instead of retyping it.
  • Digitize a printed letter, recipe or book page.
  • Extract a table or reference number from a photographed receipt or invoice.
  • Grab text from a slide or whiteboard photo after a meeting.
  • Pull quotes from images on social media into an editable document.

Frequently asked questions

Is my image uploaded to extract the text?
No. Recognition runs on your own device using a WebAssembly build of the Tesseract OCR engine. Documents, IDs, letters and screenshots never touch a server.
Why is there a one-time download?
The OCR engine (≈5 MB) and a trained data pack for each language you select (≈1–3 MB each) have to reach your browser once. You are told the size before anything downloads, and everything is cached locally — later runs work instantly, even offline.
Which languages are supported?
Twenty: English, Spanish, Portuguese, French, German, Italian, Dutch, Polish, Ukrainian, Russian, Turkish, Arabic, Hindi, Indonesian, Vietnamese, Japanese, Korean, Chinese (Simplified and Traditional) and Estonian. You can select up to three at once for mixed-language images.
How accurate is it?
On clean screenshots and scans, very accurate. Photos work best when the text is sharp, well-lit and roughly horizontal. A confidence score is shown with every result, and the output is editable so you can fix stray characters before copying.
What can I do with the result?
The extracted text appears in an editable box — copy it with one click or download it as a .txt file named after your image.
Can it read handwriting?
No — like most OCR engines it is built for printed and on-screen text. Neat block capitals sometimes work, but cursive handwriting is not supported.