OCR — scans to text
Recognize text in scanned PDFs and images.
Formats: JPG, PNG, WebP, GIF, BMP or PDF in · TXT out · Limits: Up to 20 files, 25 MB each · up to 30 PDF pages
Runs Tesseract (WASM) in your browser. Language data (about 10–15 MB per language) downloads from a public CDN on first use.
ToolDock is free and shows ads. Google and its partners may use cookies to serve and measure them.Privacy Policy
Drop files here or choose files
JPG, PNG, WebP, GIF, BMP or PDF in · TXT out
Recognition settings
The recognition engine runs on your device. Language data (about 10–15 MB) downloads once from a public CDN the first time a language is used.
About ocr — scans to text
OCR turns a scan into text: the letter shapes are recognised and written out as characters you can edit, search or paste into a system. This tool runs Tesseract, a long-established open-source recogniser, inside your browser, so a scanned contract or medical letter never leaves your device — which is often the reason to run it locally rather than upload it somewhere.
How to use ocr — scans to text
- 1
Drop your files in — or tap to browse
- 2
Set any options you need
- 3
Download the result
Why use ocr — scans to text on ToolDock
Worth knowing
Runs Tesseract (WASM) in your browser. Language data (about 10–15 MB per language) downloads from a public CDN on first use.
Formats it handles
JPG, PNG, WebP, GIF, BMP or PDF in · TXT out.
Limits
Up to 20 files, 25 MB each · up to 30 PDF pages.
Processed on your device
Runs entirely in your browser — your files never leave your device.
Frequently asked questions
How accurate is it, and what hurts accuracy?
On a clean, straight, well-lit 300 DPI scan of printed text it is typically very good — small errors, mostly in unusual names and punctuation. Accuracy drops sharply with phone photos taken at an angle, skewed pages, low contrast, handwriting and decorative typefaces. Deskewing and brightening an image before running it helps more than any setting.
Why does the first run take a while?
The recognition engine and the language data are downloaded the first time you use it — roughly 10 to 15 MB per language — and then cached by the browser. Larger documents also take real time, because recognition runs on your own processor, with no server to wait for.