About copyfrom.net

copyfrom.net does one job: get text out of things that will not let you select it — photos, screenshots and PDFs — and put a clean, copyable result in your clipboard.

The important part is where that happens. Typical “free OCR” sites upload your file to a server, process it there, and ask you to trust a delete policy. We do not have that server. The OCR engine (PaddleOCR as WebAssembly) and the PDF parser (pdf.js) run inside your browser. Your file never leaves the device.

What you can do

• Whole file. Drop or paste an image, or open a PDF. Digital PDFs are read from the text layer; scanned pages fall back to OCR automatically.
• A region. After upload, the preview stays on screen. Drag a box, or click a block to auto-detect it, and extract only that area.
• Tables. Grid-like regions are reconstructed as real tables you can copy into a spreadsheet, Markdown or CSV.
• Cleanup. An optional pass joins hard-wrapped lines and stray hyphens — without rewriting your wording. Toggle it off and the raw result is still there.

Dedicated pages live atImage to Text,Screenshot to Text andPDF to Text. They share the same engine; only the copy and accepted file types change.

Why it is free

Processing runs on your machine, so we are not paying for your CPU. After the models load once, there is no upload queue and no reason to meter pages. No account, no watermark, no “3 files a day”.

The stack, briefly

Static pages in Astro, interactive bits in React. OCR uses PP-OCRv6 models through ONNX Runtime Web. Table structure and optional smart selection use additional models we host ourselves. PDFs go through pdf.js, with CMaps and standard fonts served from this domain so CJK pages render. The site is on Cloudflare Pages.

Model files and the WebAssembly runtime are downloaded from copyfrom.net, not from a random CDN. The one optional third party is Google Analytics, and only when it is configured — seePrivacy. Block it and the tools still work.

Get in touch

Bugs, odd PDFs, or a feature that should exist:[email protected]. The source is atgithub.com/bearBoy80/copyfrom.