A PDF viewer and editor that shows you what is really in the file.
pdfcer is a single Windows executable. Unzip the build folder, run it — nothing is installed.
It opens the ordinary things: reports, scans, forms, print-shop artwork. It also opens the awkward ones — the A1 site plan, the CAD export with a hundred and fifty thousand drawing operators on one sheet, the file another viewer calls damaged. Most of the work in pdfcer has gone into reading PDFs faithfully, and that shows up as files simply looking right.
The compatibility work is done against corpora of thousands of real-world PDFs from every kind of producer — not against a handful of samples.
CMYK is converted with measured data, not a formula. The PDF standard deliberately says nothing about how CMYK should look on a screen, so every program has to choose. pdfcer's choice is a 1,296-point table fitted to measured render output: solid cyan, solid black ink and every overprint of them are values somebody actually measured, not what arithmetic predicts. Arithmetic without data is off by an average of 32 shades out of 255; this table is off by about one. Every CMYK colour in the file goes through it — fills, images, and the colour the properties panel shows you — so a rectangle and a photograph of the same colour agree. No colour profile is shipped or needed, so there is nothing to license or install.
And when a file asks to be composited in ink, it is. Artwork that declares a CMYK blending space is blended on separate colorant planes — the four process inks, plus a plane of its own for each named spot, a varnish or a Pantone — with overprint and blend modes worked out there and the conversion to screen happening once, at the end. Blending in the wrong space is how a rich black quietly becomes grey.
Transparency behaves. All sixteen blend modes, transparency groups and soft masks — including the four awkward ones, hue, saturation, colour and luminosity, that are the easiest to get wrong.
Damaged files still open. When a PDF's internal index is broken, pdfcer ignores it and scans the file for the objects themselves. Over a 1,109-file corpus of real-world PDFs, 605 of the 712 files that would not load at all failed for that one reason, and 566 of them open after the rescan.
Open a big drawing and actually move around it. Panning and zooming stay smooth on sheets that make other viewers stutter, because the last good picture stays on screen — in the right place — while the next one is drawn. Small moves cost you nothing at all; a fast sweep right across a big sheet still has to catch up.
Zoom in until you can see anything. Up to a trillion percent, and the detail is genuinely there rather than a blurred enlargement. A benzene molecule drawn at true scale is legible on screen. How far it will go is a setting you own — how much performance to spend on magnification isn't pdfcer's decision to make.
Measure on a scaled drawing. Click two points, pick two lines already on the drawing, click around a shape for the whole perimeter, or click a hole for its radius or diameter. Picks snap to the real geometry rather than near it, the scale you calibrate is written into the file and read back next time, and a measurement stays editable after you place it. (Area isn't there yet, and an angle comes from picking two lines rather than from a tool of its own.)
Mark up a set for somebody else. Sticky notes, text boxes, arrows, shapes, revision clouds, freehand, highlight / underline / strike-through, and stamps. Copy and paste markup between pages and between documents with its appearance intact — and where a kind cannot be copied, it is refused by name rather than dropped.
Read other people's comments properly. Click a sticky note on the page and it opens with its author, its date and its words. The comment list filters by author, by type, and by whether a comment actually carries any text — then jumps you to it on the page.
Change what's on the page, not just what's stuck to it. Edit text in place, restyle text that is already there — font, size, bold, italic, colour — and reflow a paragraph around the change when you ask it to. Drag the nodes of a vector path. Select an object and edit its real properties — position, size, colour. Colour controls show a proper indeterminate state when a selection disagrees with itself, and refuse by name when a colour would destroy a spot ink.
Reorganise pages — across documents. Insert, extract, rotate and delete pages. Open several documents at once and drag pages between them — dragging copies them, holding Shift moves them, and the status line says which before you let go.
Redact so that it stays redacted. The content is removed, not covered with a black rectangle that a text search finds anyway. You choose where the redacted document goes, including leaving the removal armed until you save. pdfcer tells you what it could and could not read while doing it.
Fill in a form — or build one. Text fields, checkboxes, radio buttons, dropdowns and push buttons — filling them in and authoring them both, and a placed field's properties stay editable.
Find out whether a signature means anything — and sign a document. Who signed, what they covered, and whether the bytes still match — and where the answer is no, it says why. Checking the signer against Acrobat's own certificate list is there to switch on.
Print what you meant to print. A live preview beside the settings, so you can see the sheet and what you are changing at the same time — and the commit button says in words how many sheets will be clipped.
Get the content back out. Export to DXF, PNG, JPEG, SVG, EMF, plain text or form data — or copy page content straight to the clipboard as vector, which pastes into Word as shapes you can still pull about rather than as a picture of shapes. (For a PDF out, it is Save a copy — or Save a compacted copy, which rewrites it smaller.)
Three modes, so the tools match the job. Read hides the editing machinery entirely. Review gives you markup and page operations. Edit gives you everything. Panels dock, split and rearrange, and remember where you left them between sessions.
It never scribbles on your document to tell you something. No dashed outlines, red flags or "provisional" tints painted over the page. What you see while editing is exactly what you'll get when you save. When pdfcer has had to guess at something — a substituted font, a best-fit scale, an OCR result you can't see — it says so in the status bar or in a report, off the page, where it can't be mistaken for part of your document.
It would rather tell you than quietly cope. Every page that gets rendered comes back with around a hundred named counts of what had to be approximated, and the program turns those into sentences. If a colour can't be applied, a font isn't installed, a search couldn't read a region, or a page wouldn't render, you get a sentence that names the thing and what to do about it. Silence is the one failure mode that costs you a file you thought was fine.
pdfcer is pre-1.0 and under active development — new builds most days, and each release note says what changed in plain terms. It is used daily on real files by its author, which is why it gets fixed fast; it is not yet a program with years of other people's edge cases behind it.
FEATURES.md is the authoritative list, and it makes a
distinction worth knowing about: a feature is ticked only when somebody has
driven it in a running window. "The code exists and the tests pass" gets a
different mark, because it is not the same claim. Some of what is listed above
is newer than its verification; FEATURES.md marks every one of those
separately, and if one of them misbehaves that is worth a line in the issues —
it is the half of the list nobody has walked yet.
| Where to look | What it holds |
|---|---|
MANUAL.md |
the user manual — the first five minutes, every shortcut, and what to do when something goes wrong |
FEATURES.md |
what works today, and what is coming next in order |
| Releases | every build, newest first, with notes |
Open a PDF by dragging it onto the window or pressing Ctrl+O. Scroll to move, Ctrl+scroll to zoom at the cursor. The rest is on the ribbon.
pdfcer-gui is MIT — see LICENSE. That covers everything in
this repository, including the icons, which are the author's own work.
It does not cover everything inside the executable: pdfcer links its PDF engine
statically, and that engine embeds third-party font faces and data tables.
Their notices travel with every build, in full, as
THIRD_PARTY_LICENSES.md — and in the program itself under
File ▸ pdfcer ▸ About pdfcer.
The pdfcer engine — KenM76/pdfcer — which does the PDF reading, writing and rasterizing, and has its own command line. This repository is the desktop program: the window, the ribbon, the canvas and the editing.
DEVELOPING.md
is the engineering front page — architecture, the reusable egui-shell crate,
the build and packaging tooling, and the measured performance work.
RESUME.md is where
a working session starts.