Extract text from a document image

Prepare a readable image

Use a copy such as documento.jpg or documento.png. PDFs are not accepted directly: prepare individual page images and process up to 10 at a time.

Zoom in and check whether the text is readable. If it is blurred, too distant, or cut off, take a better photo. Crop uses a center crop by ratio, not a freely selected area. OCR has no perspective correction or deskew controls.

1. Load the image and language

1. Open Extract Text with a connection available if the models have not been downloaded.
2. Wait for model loading and select the image.
3. Set OCR Languages to Spanish/English, Spanish, or English, according to the content.

A Japanese interface does not enable Japanese recognition. Downloading a model does not upload the photo.

2. Process and review

1. Click Process OCR and wait for the result.
2. Find the file headers, such as === nombre ===.
3. Review O versus 0, l versus 1, digits, decimal separators, negative signs, names, accents, addresses, and email addresses.
4. Check reading order and line breaks, especially in columns.
5. Correct the result in the text panel. This does not edit the original image.

If no text appears, check the image and recognition language. Changing the interface theme does not improve recognition.

3. Export the result

Use Export text and choose a folder. The file uses texto_extraido.txt, with a counter when necessary. It includes file headers and is not a Word or PDF document.

Copy text copies the same panel. To copy only a fragment, select it with the text editing controls or adjust the panel first.

4. Keep a verifiable copy

Export before changing the selected files or using Clear, which clears the result. Keep the source image with the TXT so you can verify doubtful text. Automatic recognition can make mistakes.

For offline use, prepare the languages you need while connected and test them before disconnecting.

• OCR controls and limits
• Privacy and offline use
• Common problems