Skip to content

Image to Text

Turn a screenshot, scanned page or photo of a document into editable text — recognised on your own device, in English, French and four more languages.

Drop, paste or pick an image

JPG · PNG · WebP · AVIF — up to 50 MB

In browser

Settings

Text language

Pick up to 3. Each extra language slows recognition a little.

The first run downloads the OCR engine and a 1–3 MB model per language, then they're cached. Your image is not sent anywhere.

Everything runs in this tab. Your photos are never uploaded to a server.

How image-to-text OCR works in your browser

The tool runs Tesseract, the open-source OCR engine, compiled to WebAssembly (via tesseract.js). It works in a background worker, so the page stays responsive while it reads.

Lines first, then characters

Tesseract finds text lines and words in the image, then an LSTM neural network reads each line as a sequence of characters, using the selected language's model.

Language models matter

Picking French teaches the engine accents and words such as « déjà » or « l'œuvre ». Mixing languages works, but each extra model adds time.

Output is plain text

You get line breaks as detected, not layout: tables, columns and bold are flattened. Edit the result in the box before copying or downloading.

OCR accuracy: what helps and what breaks it

Recognition quality depends far more on the image than on the engine. The confidence score under the result is Tesseract's own estimate — useful as a warning, not a guarantee.

Source imageWhat to expectTip
ScreenshotBest case: sharp, straight, high contrastCrop out toolbars and icons first
Flatbed scanVery good at 300 dpiScan in greyscale, not 1-bit black & white
Phone photo of a pageGood if flat and evenly litShoot straight on; avoid shadows across the lines
HandwritingPoor — the model is trained on printUse block capitals, or retype short notes

Resolution and contrast

Body text should be at least about 20 pixels tall. Tiny, blurry or low-contrast text (grey on grey, text over photos) produces substitutions like 0/O and l/1.

Skew and curvature

Lines should be horizontal. Rotate sideways scans first and flatten book pages; strong curves near the spine confuse line detection.

What it won't read well

Handwriting, decorative fonts, text on busy backgrounds and very short fragments such as single digits. Check numbers, names and amounts by eye.

Private OCR: what stays on your device

Many online OCR sites upload your image to their servers. Here the recognition happens entirely in your browser tab.

What is downloaded

The OCR engine comes from our own site. Language models are fetched once from the official tesseract.js data packages on jsDelivr and kept in your browser's storage.

What is never sent

Your image and the extracted text. Invoices, IDs or medical letters can be converted without leaving the device.

Limits of a browser tool

One image at a time, up to 50 MB, shrunk to 4000 px on the long side. For multi-page PDFs, export the pages as images first.

Image to Text: questions

How do I convert an image to text for free?

Upload the image, choose the language of the text and press Extract text. When the progress bar finishes, the text appears on the right: edit it, copy it, or sign in to download it as a .txt file.

Is the image to text tool accurate?

On clean screenshots and scans it's usually close to perfect; on photos it depends on light, angle and sharpness. The confidence score flags weak results, and you should always proofread numbers.

Can it read French text with accents?

Yes. Select French (it's preselected on the French site) so accents, ligatures and apostrophes are recognised. Combine French and English for bilingual documents.

Can it read handwriting?

Not reliably. The models are trained on printed text, so cursive handwriting mostly comes out garbled. Neat block capitals sometimes work.

Why is the first extraction slow?

The engine and the language model (1–3 MB each) are downloaded once. Later runs reuse the cached files and start in a second or two.

Are my images or text stored anywhere?

No. OCR runs in your browser; the image and the result stay in the tab and are gone when you close it. We never receive either.

Other Analyze tools people use alongside this one.

All Analyze tools