Free Image to text (OCR)

Choose an image up to 60 MB (same rules as our image tools: common image/* plus .heic / .heif / .tif / .tiff when MIME is missing). Images are scaled so the longest side is at most 3072 px before OCR to balance quality and memory. Recognition uses Tesseract.js in your browser; the engine and language files load from public CDNs on first use — Safe Local Tools never receives your image bytes. Accuracy depends on sharpness, lighting, and language choice.

OCR runs entirely in this browser tab. The first run for a language downloads traineddata from the network (often a few MB). Results are best on clear, high-contrast printed text.

Drop an image here or browsePNG, JPEG, WebP, GIF, HEIC/HEIF, TIFF, and other image/* your browser can decode; .heic / .heif / .tif / .tiff also accepted when the MIME type is blank. Max 60 MB per file.

Text language (Tesseract traineddata)

Page layout (segmentation)

If results look scrambled on tables or receipts, try Single block. For scattered short text, try Sparse.

Recognized text

Loading…

How to use this tool

1
Open Image to text (OCR)
Use it for this task: PNG, JPEG, screenshots and photos to text OCR, 100+ language packs, on-device processing, no upload.
2
Choose an image file
Drag an image into the upload area or click the button, then adjust the image options shown on the page.
3
Download or copy the image result
Preview the processed image first, then download it or copy it when the tool supports copying.

Quick facts

Runs fully offline?	After language packs and the engine cache in your browser, repeat OCR can work offline. The first download needs a network.
Is the image uploaded?	No. Safe Local Tools is static hosting; pixels stay in this tab. Tesseract fetches its own engine and traineddata from CDNs you already trust for HTTPS browsing.
Handwriting or stylized fonts?	Expect mixed results — Tesseract is tuned for printed text. Clean scans work best.
Same HEIC path as image tools?	Yes — decoding reuses the same HEIC/TIFF fallbacks as the image compressor when the browser cannot decode natively.

Top use cases

PNG, JPEG, screenshots and photos to text OCR, 100+ language packs, on-device processing, no upload.
Runs fully offline?: After language packs and the engine cache in your browser, repeat OCR can work offline. The first download needs a network.
Is the image uploaded?: No. Safe Local Tools is static hosting; pixels stay in this tab. Tesseract fetches its own engine and traineddata from CDNs you already trust for HTTPS browsing.
Handwriting or stylized fonts?: Expect mixed results — Tesseract is tuned for printed text. Clean scans work best.
Same HEIC path as image tools?: Yes — decoding reuses the same HEIC/TIFF fallbacks as the image compressor when the browser cannot decode natively.

FAQ

Why is the first run slow?▾

Tesseract downloads the selected language model and WASM core on demand. Later runs reuse the browser cache when available.

Can I trust OCR for passwords or secrets?▾

Never paste secrets into untrusted sites. Here processing is local, but screen capture habits and shoulder surfing still matter — clear results when you are done.

Does this replace Google Lens?▾

No — Lens uses different models and cloud features. This page is a privacy-first, offline-friendly alternative for quick text extraction.

Why is my language missing from the list?▾

We ship a compact set of common traineddata ids. Add more languages in a future update if demand is high.

Can I OCR a PDF?▾

Not in this version — export a page as an image first, then run OCR here.

Does scaling hurt accuracy?▾

Very large photos are downscaled to a 3072 px longest edge before OCR to protect memory. For tiny text, crop tightly in an editor first.