Series: Building PdfWord — a free, no-backend PDF tools site (Part 6)
Most free PDF sites give you the same 10 tools: merge, split, compress, convert. I have those too. But this week I shipped 5 tools I couldn't find on any free PDF site — all running entirely in the browser, no uploads, no accounts.
Here's what they are and how they work.
1. Visual PDF Compare — differences glow red
Upload two PDFs (original + revised). Every page pair is rendered and pixel-compared; differences glow red, identical content fades. You get a similarity score per page plus a downloadable diff report.
The trick: render both pages to canvas with pdf.js at the same scale, walk the pixels, flag any with channel difference > 32 (ignores anti-aliasing noise), and paint a diff visualization. ~40 lines of JS. Adobe charges for this; mine's free.
2. Listen to PDF — your documents read aloud
Upload a PDF, and the browser's built-in speech synthesis reads it to you — with voice picker and speed control. Zero libraries, zero cost: it's just speechSynthesis + pdf.js text extraction, chunked at sentence boundaries so pause/resume works naturally.
3. Split by Size — beat Gmail's 25MB limit
Nobody does this: pick a max part size (10/25/50/100 MB) and the tool splits your PDF into chunks guaranteed under the limit. It adds pages one by one with pdf-lib and byte-measures every part — no estimation, no overshoot.
4. Grayscale PDF — save your color ink
One click turns every page black & white for cheap printing. Rendered via canvas grayscale(1) filter, re-embedded at high resolution.
5. Scanned PDF to Word (OCR) — the hard one
This took real work: Tesseract.js running in the browser via WebAssembly. Scanned pages are rendered to images, OCR'd page by page, and assembled into an editable .docx.
The gotcha that cost me 3 failed QA runs: Tesseract.js v5 requests eng.traineddata.gz by default — a file that doesn't exist unless you create it. The fix was embarrassingly simple: ship both eng.traineddata and the .gz version. The worker gunzips transparently.
And the privacy angle no other free OCR site can claim: your scans never leave your device. Every other free OCR uploads your documents to their servers. Mine can't — there is no server.
The pattern
Every one of these follows the same playbook: find a task people assume needs a backend, then prove it doesn't. pdf.js renders, pdf-lib edits, Tesseract reads, speechSynthesis speaks, JSZip bundles.
Try them: PdfWord — 34 free tools, no signup, no watermark.
What impossible-in-browser tool should I build next?
Top comments (0)