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.