Documents & scans

Screenshot to text converter

Get screenshot entities: paragraphs, labels, code lines, table text, error messages, timestamps, URLs, and visible UI copy.

Before you start

What to expect

Files are retained only for the product retention window and deleted after 24 hours unless your account policy says otherwise.
Uploads are not used to train models; AI Deep Read reads the page only to produce your requested output.
AI Deep Read costs 1 credit per rendered page and still needs human review for names, dates, legal text, and totals.
Never run commands or code copied from OCR until you compare punctuation, indentation, and identifiers with the screenshot.

Plain text copied from readable pixels, preserving line breaks where they matter for code, lists, or receipts.
Markdown-friendly structure can be selected when headings, bullets, or tables are visible.
Dark-mode screenshots, app UI labels, error messages, and copied terminal output become selectable text.
Source-image context remains important because OCR can confuse punctuation and lookalike characters.

Input: Retina PNG screenshot of a dark terminal, 42 lines of stack trace, command prompt, file paths, and one wrapped error message.
Output: TXT with 42 extracted lines and the wrapped error message joined correctly; two backticks and one lowercase l in a path needed manual correction.

Code screenshots lose indentation: Line breaks are preserved, but indentation can still shift. Review spaces, tabs, quotes, and braces before running copied code.
Dark-mode or low-contrast UI: AI Deep Read handles many dark screenshots, but tiny gray labels and antialiased text may be missed.
Rotated or skewed screenshots/photos: The model reads the supplied image; rotate and crop obvious issues first for higher accuracy.
Tables flatten into text: For numeric tables, use the image-table-to-excel workflow. Screenshot-to-text is optimized for copyable text, not spreadsheets.

Command line

ocrmypdf

bash

python -c 'from PIL import Image; Image.open("screenshot.png").save("screenshot.pdf")' && ocrmypdf screenshot.pdf screenshot-ocr.pdf

Creates a PDF text layer first; you still need pdftotext or manual cleanup to get plain text.

pdftotext -layout

bash

pdftotext -layout screenshot-ocr.pdf screenshot.txt

Useful for local layout-preserving extraction after OCR.

Can it read text from a screenshot copied from my clipboard?

Yes when pasted or uploaded as an image format. Use the highest-resolution capture instead of a compressed chat image.

Is this good for screenshots of code?

It can recover code text for reference, but OCR is not safe for execution without manual review of symbols and indentation.

Can it keep tables as tables?

Use image-table-to-excel for tables. This workflow focuses on readable text output.

Will it read tiny mobile UI text?

Often, but tiny text is less reliable. Crop the relevant area and upload the original-resolution screenshot.