Privacy

Last updated: 13 August 2026

Short version: your images, screenshots and PDFs are never uploaded. Extraction, table recognition and region selection all run on your device. This page is the long version.

Your files stay in the browser

When you extract text on copyfrom.net:

• Images and screenshots are read by PaddleOCR (PP-OCRv6) compiled to WebAssembly. Inference runs in your browser via ONNX Runtime — not on a server we operate.
• PDFs are parsed by pdf.js in the same tab. Pages with a usable text layer are read directly; scanned or broken pages are rendered to a canvas and OCRed locally. You can also force OCR on a whole PDF.
• A selected region is cropped from the on-screen preview and run through the same local pipeline. Smart click uses a self-hosted SlimSAM model, also in the browser.
• Tables are reconstructed locally (geometry on digital PDFs; layout/structure models on images and scans).

The only network requests the tools need are to fetch model files, fonts/CMaps and the WebAssembly runtime —from this domain. File bytes are never attached to those requests. After the first load you can go offline and keep extracting.

Check it yourself: open the Network tab while you extract, or disconnect after the page has loaded.

What we store on your device

• Theme. Light/dark preference is saved in localStorage undercopyfrom-theme. That never leaves the browser.
• Models. Your browser may cache the OCR and related model files like any other static asset. They contain no user data.

We do not set a first-party cookie for the product itself. There is no account and no session.

What we measure (Google Analytics 4)

We load Google Analytics 4 to see aggregate usage — page views and a small set of product events — so we know which tools people actually use. Events are sent throughgtag and are limited to a fixed whitelist in our code:

• extract_start, extract_success, extract_error,region_extract, copy_result, copy_table,download_result, cleanup_applied, theme_toggle

Those events may include only coarse tags, for example:

• which tool page you were on (image / screenshot / PDF);
• how the file arrived (upload, paste, drop) and whether it was an image or a PDF;
• page count, how many pages used OCR, table count, duration in milliseconds;
• error class (unsupported, empty, engine) — never the file name;
• region method (drag or point); table copy format (Excel / Markdown / CSV);
• theme (light or dark).

They never include your text, images, PDFs, file names, or the pixels of a selected region.

GA4 is a third-party service. When it is on, Google may set its own cookies or identifiers. We configure the tag with IP anonymization and without Google Signals or ad personalization. SeeGoogle’s privacy policy for how Google processes that data.

If you block analytics (browser setting, extension, or content blocker), every tool on this site keeps working.

What we do not do

• No accounts, so no email, password or profile.
• No advertising pixels and no selling of data — we never receive your files to sell.
• No server-side copy of what you extract.

Hosting

The site is static files on Cloudflare Pages. Like any host, Cloudflare sees ordinary HTTP metadata (IP address, user agent, the URL you requested) when it serves those files. That is delivery and abuse-prevention logging, not a copy of your document.

Children

The tools are general-purpose utilities. We do not knowingly collect personal information from children. There is nothing to create an account with.

Changes

If this policy changes in a material way, we will update the date at the top of this page. The current text is always at /privacy/.

Questions

For a privacy question or a correction, write tohello@copyfrom.net. Also see About and Terms of Use.