About PDFBucket
One developer, one frustration with cloud tools leaking private files, and a weekend building something better. Here's the full story.
Jeeva
Founder & Solo Developer
Full-stack developer based in Tamil Nadu, India. I build tools that solve real problems I run into — and then make them free for everyone else who has the same problem.
Why I built PDFBucket
In March 2024, I was helping a client prepare a set of financial documents — salary slips, bank statements, and a signed contract — and needed to compress each PDF to meet a 2 MB email attachment limit. I went to Google, found the top three tools, and then noticed something: all of them uploaded the file to a server before compressing it. One even displayed a “Processing your file on our secure servers” message.
These were confidential financial documents. I wasn't comfortable uploading them to an unknown server. So I opened a browser tab, wrote a quick Canvas API compression script, and it worked instantly — no upload, no server, no privacy exposure.
That weekend I turned the script into a proper tool. Then added an image resizer. Then a format converter. By June 2024, it was a suite of 7 tools. I kept them free and open-source because if I needed this, other people definitely do too.
The technical guarantees
These aren't marketing phrases — they're specific architectural decisions you can verify in the source code.
ONNX Runtime Web
The Background Remover runs a U²-Net segmentation model via ONNX Runtime Web. The model file is ~10 MB and caches locally after first load — no cloud AI API is called.
ffmpeg.wasm
MP4-to-MP3 uses a full WebAssembly port of FFmpeg. The same open-source engine used by VLC — running locally on your CPU, not a cloud transcoding service.
Tesseract.js
OCR uses Tesseract.js v4 with an LSTM neural network model. Google's open-source OCR engine, compiled to WebAssembly, running in a browser Web Worker.
pdf-lib + mammoth.js
PDF merging uses pdf-lib; DOCX-to-PDF uses mammoth.js for parsing. Both are pure JavaScript — no native dependencies, no server calls.
Canvas API
Image compression, resizing, format conversion, and colour palette extraction all run on the browser's native Canvas API — no third-party image processing library needed.
Open source
All source code is public. You can inspect every library call, verify no hidden network requests are made, and run the whole thing locally if you prefer.
Don't trust — verify
Every claim on this page is verifiable in the source code. Open DevTools on any tool page and watch the Network tab — you'll see zero file uploads. Or read the code directly on GitHub.
View Source on GitHub →