Local OCR (Image to Text)
Extract text from an image — a photo of a document, a screenshot, a scanned page — using a real OCR engine that runs entirely on your device via WebAssembly. Works best on clear, reasonably high-contrast printed text; a confidence score is shown alongside the result so you're not left guessing how much to trust it. Nothing about your image is ever uploaded — the OCR model (~8MB) downloads once and every image after that processes completely offline.
About this tool
This tool extracts text from an image — a photo of a document, a screenshot, a scanned page — using a real OCR (optical character recognition) engine that runs entirely on your device via WebAssembly. It's genuinely useful for sensitive documents you don't want to send to a server just to get the text out of them: a receipt, an ID, a handwritten note photographed for reference. It works best on clear, reasonably high-contrast printed text; handwriting, stylized fonts, low-resolution scans, and heavily skewed photos will all recognize noticeably worse — a confidence score is shown alongside the result so you're not left guessing how much to trust it. Nothing about your image is ever uploaded — the OCR model itself (~8MB) downloads once from Tuliq and every image after that processes completely offline.
What you can do
- Extract text from a photo, screenshot, or scanned document
- Runs a real OCR model on your own device, not a server
- Copy the result or download it as a .txt file
- Your image is never uploaded — recognition runs entirely in your browser
How to use
Step 1
Upload an image
A photo of a document, a screenshot, or a scan — clear, reasonably high-contrast text works best.
Step 2
The model reads it
Text recognition runs entirely on your device via WebAssembly.
Step 3
Copy or download
Get the recognized text as plain text, with a confidence indicator.
Example
A photo of a printed page:
Result
Extracted text, with a confidence score like "92% (high confidence)" shown alongside it.
Clear, well-lit, high-contrast text typically recognizes at high confidence; a blurry photo or unusual font will score lower — worth a manual check before trusting it fully.
How your data is handled
This tool is labeled 🔵 Local AI: your image is read directly using your browser's File API and text recognition runs entirely on your device via a local WebAssembly model. It's never uploaded anywhere. The one network request this tool makes is downloading the model files (~8MB total) once, which are cached by your browser afterward — every image you process after that runs completely offline.
Frequently asked questions
Does it work on handwriting?+
Not well — this model is trained for printed text. Handwriting, especially cursive or stylized handwriting, will recognize poorly or not at all. It's built for photos, screenshots, and scans of printed or typed text.
What languages does it support?+
English only for now. Supporting more languages means downloading additional trained-data files per language, which isn't built yet.
What does the confidence score mean?+
It's the OCR engine's own estimate (0-100%) of how sure it is about the recognized text, shown as a plain high/medium/low label. Low confidence usually means blurry input, an unusual font, or a heavily skewed photo — worth double-checking the result by eye.
Is my image uploaded anywhere?+
No — it's read directly in your browser and text recognition runs on your device via a local WebAssembly model. It never leaves your device, and nothing is retained once you leave the page. The only network request is a one-time model download (~8MB total), cached by your browser afterward.