Desktop software
Spectra PDF
A free, open-source PDF workbench for Windows.
What it is
Spectra PDF is a free, open-source PDF workbench for Windows. A menu bar, customizable toolbar, and document tabs sit over a continuous reading view, with a navigation pane on the left and twenty-four tools that open in a resizable pane on the right, so the document never leaves the screen. The keymap is checked against the industry-standard editor's published shortcut table, so existing muscle memory works. No ads, no telemetry, no upsells.
The tools cover the whole everyday workflow: organizing pages, commenting, filling and signing forms, preparing new ones, redaction that actually removes content, scan and OCR, document compare, encryption, optimization, repair, watermarking, headers and footers with Bates numbering, page labels, attachments, portfolios, layers, an accessibility checker, print preflight, link management, and export to text, Word, RTF, ODT, HTML, or page images.
Every whole-file operation also works from the command line. The same engine runs headless with identical results, so batch work is scriptable.
Why I built it
PDF tooling on Windows is usually a subscription, a web service that wants your documents uploaded, or a free utility that does exactly one thing. I wanted the full everyday workflow as ordinary desktop software that respects the machine it runs on.
It started as a small tool for splitting and merging. It spent a few weeks under a different name, until I found an established open-source project already using it — one I like — and renamed rather than muddy theirs. Spectra PDF is what it was called first and what it is called now.
How it works
Tauri v2 with a React front end. An embedded Python engine does the document work, and a vendored upstream Ghostscript build handles compression, color conversion, PDF/A, printing, and PostScript conversion. The reading view virtualizes rendering, so thousand-page documents scroll smoothly. Page edits stage in memory and commit atomically, with multi-level undo across both staged edits and applied operations. It also supports .pdfx — Alexandros Gounis's open format that stores several documents in one ordinary, fully compatible PDF, which reopens as separate strips. Good idea, credited.
OCR runs entirely offline, including a batch mode that walks a folder tree and returns a mirrored copy with every scanned PDF made searchable. Files that are already searchable copy through untouched, problem files are reported, and the originals are never modified.
It installs the way an endpoint administrator would want: an NSIS installer with silent modes and enterprise policy support, file associations, Explorer context-menu entries, and update notifications. The app tells you a release exists; it never downloads or installs one on its own.
Editing text that was never meant to be edited
A PDF does not contain paragraphs. It contains glyphs at coordinates. Editing text in place means reconstructing the paragraph nobody stored, changing it, and putting it back in the document's own font so the result is indistinguishable from what was there before.
That means true rewrap when a line grows, kerning that matches the original typesetting, splitting and merging paragraphs, and restyling a range — size, color, family, bold, italic, and real OpenType small caps and stylistic alternates — without leaving artifacts. Right-to-left scripts raise the difficulty again: Arabic, Hebrew, Persian, and Urdu have to reflow and author like any other text, with cursive letters joined correctly. Chinese, Japanese, and Korean too.
The details compound. Combining accent marks have to compose onto their letters instead of stranding beside them. Ligatures should form where the typeface carries them, while the words still copy, search, and extract as ordinary letters. Older CJK encodings — Shift-JIS, EUC, Big5, GBK — mix one- and two-byte characters on the same line and have to open for editing rather than being declined. And when the font picker lists what is installed on the machine, it has to leave out faces whose license forbids embedding, then say how many it left out rather than letting someone hunt for a missing font.
This is where most free PDF tools stop, and it is the work that separates a viewer with extras from an editor.
Status
Actively developed, MIT-licensed, on GitHub, with installers on the releases page. The bundled Ghostscript is upstream AGPL-3.0, vendored unmodified.