MarkdownPDF

Image to Text (OCR)

Get the text out of a photo, a screenshot or a scan. Seven languages including Arabic, free, and the image never leaves your browser — paste it with Ctrl+V if it is already in your clipboard.

Drop images here, or browse

Processed on your device — never uploaded · or paste a screenshot with Ctrl+V

Why do this locally?

Think about what people actually run through an image-to-text tool: a screenshot of a conversation, a photograph of a prescription, an ID card, a bank statement, a page of a book they do not own the rights to. Every online OCR service receives a copy of that picture. Here the recognition engine is downloaded into your browser and the image is read on your own machine — see how it works for the two checks that prove it.

Getting good results

OCR quality depends far more on the picture than on the engine. Hold the text upright and parallel, fill the frame, use even light, and avoid the shadow of your own phone. A flat scan at 300 dpi beats a photo every time. The full checklist, including the error patterns worth proofreading for, is in the guide to extracting text from scans.

Whole documents

For a multi-page PDF, use OCR a PDF to keep the document and make it searchable, or PDF to Markdown to pull the content out with its headings and lists intact.

Frequently asked questions

Which image formats can I use?

PNG, JPEG, WebP, GIF, BMP and TIFF. You can also paste a screenshot straight from the clipboard with Ctrl+V — usually faster than saving it first.

Which languages does it read?

English, French, Spanish, German, Portuguese, Italian, Arabic. Choose the language before dropping the image: recognition leans on a language model, so the right choice matters, particularly for accented or Arabic text.

Is my image uploaded?

No. The OCR engine runs in your browser and is served from this site, not a third-party CDN. The picture stays on your device — which matters for screenshots of private conversations, ID documents or medical results.

Can it read handwriting?

Not reliably. This engine is trained on printed text. Neat block capitals sometimes come through; ordinary handwriting does not. Recognising handwriting is a different technology altogether.

The text came out wrong — what should I change?

Almost always the picture rather than the settings: get the text upright and filling the frame, avoid shadows and glare, and prefer a sharp photo over a zoomed-in crop. Then check the language selector. Tiny text in a low-resolution screenshot is the hardest case.

Can I read several images at once?

Yes — drop up to 20. They are processed one after another and the results are concatenated with each file name as a separator, so you can tell which text came from which image.

Limitations

Guides