Free PDF Tools Without Uploading: Compress, Sign, and OCR — Without Your Documents Leaving Your Computer

Published: 28 July 2026 · 9 min read

Type "free PDF compressor" into Google. The first 10 results all promise the same thing: drag in your PDF, get a smaller one back, free. What none of them prominently disclose: your PDF is getting uploaded to a server somewhere. The company behind the free tool is paying for server infrastructure, and they're recouping that cost somehow — usually by mining the documents you upload for data they can monetize, or by upselling you to a paid plan after you've invested time in their workflow.

For most PDFs, this matters less than people think. A restaurant menu PDF isn't sensitive. But what about a signed contract? A tax document? An NDA? A medical record? A legal brief? Once you upload those to a "free PDF tool," you've lost control of them. The company's privacy policy — if you read it — almost certainly says they can process, store, and analyze uploaded content. Some explicitly claim rights to use uploaded documents for product improvement, which in the AI era increasingly means "training data."

This guide covers what PDF tools you actually need, which ones process files locally in the browser, and how to verify the privacy claims for yourself.

The PDF tool landscape: what's actually useful

There are dozens of PDF tools advertised. Most people only need five:

  1. Compress — make a PDF smaller for email or web upload
  2. Sign — add a signature or fill form fields
  3. Merge — combine multiple PDFs into one document
  4. Split — extract specific pages from a larger PDF
  5. OCR — make scanned documents searchable by recognizing text in images

Everything else — PDF to Word conversion, PDF editing, PDF annotation, PDF password protection — is either a variation of these five or handled better by dedicated software. Converting PDF to Word has been a solved problem since the 2000s but still generates enormous search volume because people keep looking for free alternatives to Adobe Acrobat ($19.99/month). The answer is usually LibreOffice (free, local, handles most PDFs reasonably well) or just exporting from the original program.

PDF compression: server vs browser approaches

PDF compression works by applying several strategies:

  • Image recompression: Most of a PDF's size is usually embedded images. Recompressing those images to JPEG with a lower quality setting can reduce PDF size by 50-90% with negligible visual difference.
  • Font subsetting: Instead of embedding entire font files (which can be megabytes each), embed only the characters actually used in the document.
  • Object deduplication: PDFs often contain duplicate objects — the same image used on multiple pages, the same font data, identical metadata blocks. Removing duplicates saves space.
  • Metadata stripping: Remove author info, editing history, embedded thumbnails, and other cruft that bloats file size without adding value.

Server-side PDF compressors use libraries like Ghostscript (AGPL, extremely capable, 25+ years of development) to do all four. Browser-based compressors typically use pdf-lib, a JavaScript library that handles the PDF file structure manipulation but can't do Ghostscript-level image recompression. The result: server-side compression usually achieves better ratios (60-80% reduction vs 30-50% for browser-based), but at the cost of uploading your document.

The pragmatic approach: If the PDF is non-sensitive (public report, ebook, brochure), use whatever tool gives the best compression. If the PDF contains anything confidential — legal documents, financial records, personal information — the 10-20% difference in compression ratio isn't worth the privacy tradeoff. Use a browser-based tool and accept the slightly larger file.

PDF signing: the one tool you should never use a cloud service for

A signed PDF with your actual signature is one of the most sensitive documents you'll ever handle digitally. A signature on a blank check. A signed employment contract. A medical consent form. Uploading these to a cloud PDF signing service means a third party now has:

  • Your full legal name
  • Your handwritten signature (which can be extracted as an image)
  • The contents of whatever document you're signing
  • Metadata including when you signed it and your IP address

Browser-based PDF signing tools exist that do all processing in local memory: load the PDF, render it to a canvas or use pdf-lib to manipulate the file structure, let you draw or upload a signature image, flatten the signature onto the page, and produce a signed PDF. The document never leaves your computer. SignPrivately does this — upload the PDF, sign it, download the signed version. Zero server interaction.

For legally binding signatures, the standard is higher. eIDAS in the EU and the ESIGN Act in the US define what makes an electronic signature legally valid. A flat image of your signature pasted onto a PDF is technically an "electronic signature" under US law and is legally binding for most purposes. But for high-stakes documents, you may need a "digital signature" with cryptographic proof of identity (DocuSign, Adobe Sign). Those services charge for that cryptographic infrastructure — and for good reason. The tradeoff is: they have your document, but they also provide non-repudiation that a flat image signature doesn't. Choose based on what the signature needs to prove.

OCR: the genuinely hard one

Optical Character Recognition — making a scanned image of text into actual searchable, selectable text — is the hardest PDF task to do in a browser. The state-of-the-art OCR engines are Tesseract (open source, maintained by Google) and various commercial offerings from ABBYY, Azure Cognitive Services, and Google Cloud Vision. Tesseract runs in the browser via a WebAssembly port, but it's slower than native Tesseract and the accuracy is worse because the WebAssembly version uses a simplified neural network component.

For English text on clean scans, browser-based Tesseract is adequate — 95-98% accuracy on 300 DPI scans. For handwriting, non-English languages, low-contrast text, or anything scanned below 200 DPI, server-side OCR is significantly better. If you need OCR for sensitive documents, your best option is to install Tesseract locally (brew install tesseract on Mac, apt install tesseract-ocr on Linux) and run it from the command line. It's free, it's fast, and your documents stay on your machine.

How to check if a PDF tool is uploading your files

Same technique as checking background removal tools. Open browser DevTools (F12), go to the Network tab, check "Preserve log," and process a PDF. Look for:

  • Any POST or PUT request with a payload similar in size to your PDF
  • Any request to a domain that isn't the CDN the tool loads libraries from
  • WebSocket connections (some tools stream the PDF for "real-time processing")

If you see a request sending multiple megabytes to api.pdftool.com, your document is on their server. If the only traffic you see is CDN library loads (jsDelivr, cdnjs, unpkg) and the tool still works, it's processing locally.

The acid test: after the tool's libraries finish loading, turn off your WiFi. If the tool still works, it's local. If it shows a spinner and then fails, it was uploading.

The practical PDF workflow that keeps everything local

For most day-to-day PDF work, here's a workflow that uses zero cloud services:

  1. Viewing/reading: Your browser's built-in PDF viewer, or your OS's default PDF reader. No upload needed — the PDF opens from your local filesystem.
  2. Compressing: Browser-based compressor using pdf-lib for basic compression. For aggressive compression of non-sensitive documents, Ghostscript CLI (gs -sDEVICE=pdfwrite -dPDFSETTINGS=/ebook -dNOPAUSE -dQUIET -dBATCH -sOutputFile=output.pdf input.pdf) reduces most PDFs by 70-80%.
  3. Signing: Browser-based signing tool. Upload nothing. Flatten the signature and download.
  4. Merging/splitting: pdf-lib handles both operations in the browser. For command-line work, pdftk (free, cross-platform) has been the standard since 2004 and still works perfectly.
  5. OCR: Tesseract CLI for non-sensitive scans. For sensitive documents, scan at 300+ DPI and use local Tesseract with the appropriate language pack.
  6. Editing: LibreOffice Draw opens PDFs for editing. It's not perfect — complex layouts sometimes shift — but for fixing a typo or updating a date in a PDF, it works. For heavy PDF editing, you probably want the original source document anyway.

None of these steps require an internet connection after initial software installation. None of them involve uploading to a server you don't control. None of them cost anything. The tradeoff is convenience — you have to know which tool to use for which task, and you don't get the slick "drag, click, done" experience of an all-in-one cloud service. Whether that tradeoff is worth it depends entirely on what's in your PDFs.

Sign PDFs in your browser — no uploads, free →