Extract text from photos and scans

Useful for students, archivists and office teams who need to recover wording from receipts, notes or scanned documents without retyping.

Turn photos and scanned pages into editable text you can copy, search and reuse, while preserving line breaks for a cleaner result.

How to use

Choose a photo or scan in PNG, JPEG, WebP or TIFF format that contains the text to recognize.
Select the document language from the available options to guide recognition.
Start the extraction and let the calculation server process the image to produce text.
Copy the resulting text, proofread it and fix recognition errors before using it.

Example

An archivist photographs a typed letter. They choose French and English, run the extraction, then compare the text with the original to correct proper names and numbers before adding it to an inventory.

Limits and formats

Quality depends on sharpness, lighting and how straight the photographed document is.
Handwriting, decorative fonts and complex tables are often poorly recognized.
The produced text must be proofread: the tool may confuse similar characters or skip lines.
One image is processed at a time, and only PNG, JPEG, WebP, TIF and TIFF formats are accepted.

Where processing happens

Recognition is a heavy operation performed on the tool calculation server. The file is uploaded, processed in memory without disk writing on the gateway, then deleted at the end. The file and its result are kept for at most 15 minutes by default, then removed. Neither the name nor the content is logged, and no third party is involved. This keeps the operation isolated from other users.

Frequently asked questions

Which languages can I recognize?

The interface offers French and English together, then French, English, German, Spanish and Italian separately. Choose the document main language to improve recognition. A mix of languages may require more corrections.

Why does the extracted text contain errors?

Optical recognition depends on image quality. Blur, glare or low contrast makes analysis harder. Similar characters, columns and stamps can also be misinterpreted. Careful proofreading remains necessary.

Can I extract text from a confidential document?

The file is sent to the calculation server for the operation and kept for at most 15 minutes by default, then deleted. For highly sensitive data, it is better to avoid this type of upload or mask confidential passages before processing.