How pdf-workshop works

pdf-workshop is a free set of 99 PDF tools and a growing set of image tools that run entirely in your web browser. There is no server that receives documents. When you open a tool, the page downloads the code it needs — including full document engines compiled to WebAssembly — and does the work on your own device. The file you chose is read from your disk into the browser's memory, changed there, and written back to your downloads. At no point is it transmitted anywhere.

Why no server

Most online PDF and image tools upload your file, process it on their machines and send a result back. That is convenient for them and a risk for you: contracts, medical records, ID scans, family photos and financial statements sit on someone else's disk, at least for a while, subject to their retention policy and their security. pdf-workshop removes the question. Because there is no upload endpoint, there is nothing to leak from. It also means the tools work on corporate networks that block uploads, and that there is no account, no quota and no per-file fee.

The engines

Serious PDF work needs serious software. Rather than reimplement it, pdf-workshop runs established open-source engines inside the browser. Each is loaded only when a tool needs it, so the home page stays light.

pdf-lib and PDF.js
Reading, assembling, rendering and writing PDF files: merging, splitting, rotating, page numbers, watermarks, signatures, form filling and previews. Merge PDF, Split PDF, Sign PDF
qpdf
The reference PDF transformation library, compiled to WebAssembly: AES-256 encryption and decryption, permissions, linearisation and repair. Encrypt PDF, Decrypt PDF, Repair PDF
Tesseract OCR
Text recognition in English and Hindi, running as WebAssembly in a worker. Used to make scans searchable and to find the words in a scanned page before editing them. OCR PDF, Edit Scanned PDF Text
PaddleOCR detection via ONNX Runtime
A neural text-detection model that locates lines of text in a scanned image so recognition can be aimed precisely, especially on noisy scans and Devanagari text. Edit Scanned PDF Text
MODNet via ONNX Runtime
A portrait matting model that predicts which pixels of a photo are the person. It lets the passport tool swap the background for a plain colour, entirely on your device. Passport Photo Maker
LibreOffice
The full office suite compiled to WebAssembly, loaded on first use. Converts Word, Excel, PowerPoint, OpenDocument and RTF files to PDF with the same layout engine as the desktop application. Word to PDF, Excel to PDF, PowerPoint to PDF
The browser's own image engine
Decoding, drawing and re-encoding JPG, PNG and WebP pictures with the Canvas support built into every modern browser. Nothing extra is downloaded. Each picture is redrawn, which also drops location and camera details stored inside it. Compress Image, Resize Image, Convert Image, Rotate & Flip Image
pdf-oxide
PDF/A conversion with font embedding and an sRGB output intent, for archives that must open the same way in twenty years. PDF to PDF/A

What happens to a file, step by step

  1. 1You choose or drop a file. The browser reads it from disk into memory; nothing is sent.
  2. 2The tool loads the engine it needs — WebAssembly modules downloaded from the site, cached by the browser after the first time.
  3. 3The engine does the work in your browser tab, or in a worker thread beside it so the page stays responsive.
  4. 4The result is written to a new file in memory and offered to you as a download, or shown in the preview so you can apply another tool to it.
  5. 5When you close the tab, the memory is released. There is no copy anywhere else.

Two things worth knowing

It needs an internet connection to load, not to work.
The page and its engines are downloaded from the site. Once a tool has loaded, the processing itself makes no network requests. There is no service worker yet, so the site does not run fully offline; that would be a true claim only once one ships.
Big jobs need a capable device.
Because the work happens on your machine, its memory is the limit. Converting a large office document with LibreOffice, or OCR on a long scan, is heavier than merging two PDFs. A recent laptop or phone handles everyday files comfortably.

The open-source projects behind all of this are listed on the credits page. Every tool is in the directory.