A PowerPoint file is a zip archive of XML describing where everything sits. That makes converting it to PDF far more tractable than converting a Word document, because a slide records absolute positions for every shape — there is no re-flow to compute, no page-break algorithm to reimplement.
What it does not record is the fonts. And that is where deck conversion actually goes wrong.
The font problem, and what a metric substitute fixes
A .pptx names the fonts it uses but does not usually contain them. Calibri is Microsoft's font, licensed with Office; a converter that is not Office cannot ship it. So the text has to be drawn in something else, and the choice of something else decides whether your slides survive.
Fonts differ in letter widths. Substitute a font whose letters are wider and a line that fitted on one row now wraps onto two, pushing everything below it down and out of its box. That is the familiar failure: a deck that converts with text overflowing its shapes.
The fix is a metrically compatible substitute — a different typeface deliberately built to have identical letter widths, so every line breaks in exactly the same place. Carlito is that font for Calibri; Helvetica already has Arial's widths.
This was checked against a real conversion rather than taken on the font project's word: on a 24-paragraph test deck, Carlito broke all 14 Calibri paragraphs in exactly the same place as real Calibri, where a standard PDF font got 11 of the 14 wrong. The letters are drawn slightly differently — that is a visual difference, not a layout one.
What cannot be drawn, and when you should be told
Charts, SmartArt, tables, embedded objects, custom shapes, and pictures in formats a PDF cannot carry — EMF and WMF among them — are beyond what a converter can render from the XML alone. Every converter that is not PowerPoint itself has this boundary somewhere.
What differs between tools is when you find out. The usual experience is to convert, download, and discover the gap while scrolling — or worse, not discover it, because the missing chart left blank space that reads as deliberate design.
Filum lists what it cannot draw before you convert, naming each item by slide number, so you can decide whether the PDF is worth making at all. Anything it cannot draw becomes a labelled box on the page saying so, never a blank space you would mistake for the layout.
Page size, scripts, and the older format
The output page is your deck's own size, taken from the slide size stored in the file — not a letter or A4 default. A standard 10 by 7.5 inch presentation produces a 720 by 540 point page. If a presentation states no usable size, the tool falls back to 10 by 7.5 inches and says so first.
Script coverage is a real limit and it is stated rather than discovered. The fonts cover Latin script and Western European accents, not Chinese, Japanese, Korean, Cyrillic, Greek, Arabic, Hebrew or Devanagari. A deck in those scripts is refused with a message naming the problem, rather than converted into a PDF of blank space.
The older .ppt format is not supported. The tool inspects the file itself rather than trusting its name, so a .ppt renamed to .pptx is caught and explained instead of producing a broken PDF — the same check that catches password-protected files, which use the same container. Open it in PowerPoint and use Save As to produce a .pptx.
Why this one runs on your device
Deck conversion is normally a server operation, because the standard approach is to run a copy of an office suite on a machine somewhere and print to PDF. That works, and it means your presentation is uploaded.
Filum converts in your browser instead, by reading the XML directly. Your deck never leaves your computer and never reaches a server, and you can confirm that in your browser's Network tab.
A deck is often the most confidential file a person owns — the board pack, the unannounced product, the reorganisation plan. It is a strange category of document to hand to a free web service, and it is the main reason this conversion was worth building the harder way.