PaperOtter is a desktop app with 19 document tools, all running on your own machine. Compress a PDF to a size you name, merge, split, rotate, sign, redact and repair, and convert PDFs or Word files into Markdown, HTML or JSON with headings and lists intact, which makes them far cheaper to feed to an LLM. Nothing is uploaded: there is no network code in the document paths at all. Free, open source, offline and no account, for Mac, Windows and Linux.
PaperOtter started with something small and annoying: I had PDFs that were too big to attach to an email. The fix everyone uses is to drag the file onto a free website, click compress, and wait. It works, and it also means handing your document to a stranger's server to make it smaller.
So I built a compressor that runs on my own machine. Then I kept finding other things I wanted from the same files, and it grew into 19 tools.
The three that changed how I actually work:
Compress to a real target size:
Not a vague quality slider: say "get this under 2 MB" and it cascades through presets until it gets there.
PDF to Markdown or HTML, with the images left out:
I feed a lot of documents to LLMs, and paying tokens for pictures I do not need is wasteful. This converts on-device, keeps the headings, paragraphs and lists intact, can drop the images, and can split a long document into one file per chapter using the PDF's own bookmarks. The document never leaves the machine on the way.
Images into a PDF, and pages back out as images:
Both directions, no website in between.
Around those sit the ordinary chores: merge, split, rotate, crop, reorder pages, page numbers, watermarks, signing, redaction that actually removes the content, and repair for files other tools refuse to open.
Two things I would want to know if I were reading this:
It has no network code:
Not "we do not look at your files": there is no HTTP client anywhere in the document paths. Block it in your firewall and every tool still works. That is a checkable claim, and the source is public.
It needs nothing installed:
18 of the 19 tools are compiled into a 24 MB binary. We removed Ghostscript entirely, which was the last AGPL dependency; the bundle went from 52 MB to 24 MB, and the Rust compressor that replaced it produces smaller files than Ghostscript did on our test corpus.
Two honest limitations, stated here rather than discovered later: OCR is macOS-only today because it uses Apple's Vision framework instead of bundling an engine, and Windows and Linux are planned. The nine interface languages were machine-translated, and the README says so.
Hi Product Hunt 👋
PaperOtter started with something small and annoying: I had PDFs that were too big to attach to an email. The fix everyone uses is to drag the file onto a free website, click compress, and wait. It works, and it also means handing your document to a stranger's server to make it smaller.
So I built a compressor that runs on my own machine. Then I kept finding other things I wanted from the same files, and it grew into 19 tools.
The three that changed how I actually work:
Compress to a real target size:
Not a vague quality slider: say "get this under 2 MB" and it cascades through presets until it gets there.
PDF to Markdown or HTML, with the images left out:
I feed a lot of documents to LLMs, and paying tokens for pictures I do not need is wasteful. This converts on-device, keeps the headings, paragraphs and lists intact, can drop the images, and can split a long document into one file per chapter using the PDF's own bookmarks. The document never leaves the machine on the way.
Images into a PDF, and pages back out as images:
Both directions, no website in between.
Around those sit the ordinary chores: merge, split, rotate, crop, reorder pages, page numbers, watermarks, signing, redaction that actually removes the content, and repair for files other tools refuse to open.
Two things I would want to know if I were reading this:
It has no network code:
Not "we do not look at your files": there is no HTTP client anywhere in the document paths. Block it in your firewall and every tool still works. That is a checkable claim, and the source is public.
It needs nothing installed:
18 of the 19 tools are compiled into a 24 MB binary. We removed Ghostscript entirely, which was the last AGPL dependency; the bundle went from 52 MB to 24 MB, and the Rust compressor that replaced it produces smaller files than Ghostscript did on our test corpus.
Two honest limitations, stated here rather than discovered later: OCR is macOS-only today because it uses Apple's Vision framework instead of bundling an engine, and Windows and Linux are planned. The nine interface languages were machine-translated, and the README says so.
Free, MIT, no account, no upsell, no pro tier.