Documents in. PDFs out. Nothing leaves your machine.

What it converts

Pictures, text and spreadsheets. Eight picture formats — JPG, PNG, HEIC, WEBP, AVIF, GIF, BMP and SVG — and four text ones: plain text, Markdown, HTML and CSV. A PDF can be dropped in too, though it is not converted: it is already the output format, so its pages are carried into a combined document as they are. Office documents are being built and are deliberately not offered yet.

Text keeps its structure, and stays text

A converted document holds real text, not a picture of text. You can search it, select it, copy out of it and let a screen reader read it, which is most of what a PDF is for and the first thing lost by converters that rasterise the page.

Markdown keeps its headings, its lists and its tables. A table is most of why anyone reaches for a converter, and flattening one into a line of text with the pipes still in it is the commonest way this goes wrong. HTML keeps its structure too, and is stripped of everything that executes or loads: script, styles, frames and remote images never reach the page.

Images inside an HTML file are dropped rather than fetched. This page is forbidden a network connection, so a remote image could not load even if it were kept, and a document with silent holes where pictures should be is worse than one that says it carries none.

A spreadsheet stays a table

A CSV becomes a real table with its first row as the heading, repeated at the top of every page — a twelve-page export whose headings appear only on page one is unreadable by page three. The delimiter is worked out from the file, so a semicolon export, which is what most of Europe produces because the comma is their decimal separator, is read without being told. Tab-separated files, which is what you get from copying out of a spreadsheet, are read the same way.

Quoted fields are read properly. An address with a comma in it, a price written with a thousands separator, a sentence with a line break inside it: all stay in one cell. Splitting a CSV on its commas, which is the quick way to do this, corrupts every file that contains one.

A table too wide for the page is split, never trimmed

Thirty columns do not fit across A4 at any readable size. The obvious answer is to let the table shrink to fit, and it does not work: measured, a thirty-column table rendered that way came out with eighteen columns and the other twelve silently gone, drawn past the edge of the paper. Nothing in the file says so — it opens, the table looks complete, and the last columns are simply not there.

So a table wider than the page is broken down its columns into blocks that do fit, each with its own heading row and a caption saying which columns it holds. Every cell survives. The cost is that a wide table is read in blocks rather than across one line, and choosing landscape or a larger page size puts more columns in each block.

The encoding is read, not assumed

A text file written on Windows is often not UTF-8, and reading it as though it were turns every accent and curly quote into a replacement character. LocalPress checks for a byte order mark, tries UTF-8 strictly, and falls back to Windows-1252 when that fails, which is where Word's apostrophes and dashes live. UTF-16 is recognised too, with or without its mark.

What it will not pretend to draw

The font that ships with the converter covers Latin, Greek and Cyrillic. A document in Chinese, Japanese, Korean, Arabic, Hebrew, Thai or Devanagari is refused before conversion, with the script named, because the alternative is a PDF of blank pages that only reveals itself when you open it. Arrows and emoji fall outside the font too, and are reported the same way. Font packs for the other scripts are the next piece of work.

SVG stays a drawing

An SVG is not turned into pixels. It is written into the PDF as vectors, so a logo or a diagram stays sharp at any zoom and prints at whatever resolution the printer has. Converters that rasterise an SVG to fit a page throw away the one property that made the format worth keeping, and the loss only becomes obvious later, on paper.

Photographs come out the right way up

A phone held sideways does not rotate the picture it records. It stores the pixels the way the sensor saw them and writes a tag saying which way up the result should be. Software that ignores that tag prints the photograph on its side. LocalPress reads it, and then measures the page against the size the picture is actually drawn at rather than the size it happens to be stored at, which is the part that is easy to get half right.

HEIC, which your browser cannot read

An iPhone photograph is usually a HEIC, and no browser except Safari can decode one. That is why the camera roll so often has to be exported before it can be used anywhere. LocalPress decodes it here, on your machine, usinglibheif compiled to WebAssembly. The decoder is 1.4MB and is downloaded only when you actually drop a HEIC, so nobody else pays for it.

It is the one piece of third-party code this page runs, and it is worth being plain about that. libheif is published under theLGPL-3.0 rather than the permissive licence everything else here uses, so the exact build is shipped unmodified as its own file, itsfull licence and notices ship with it, and its source is public. Nothing about it changes where your file goes: it runs in the same browser tab under the same policy that forbids the page a network connection.

A photograph taken with the phone held sideways is stored the way the sensor saw it, with a rotation recorded beside it in the file. That rotation is honoured, so the picture arrives on the page the way it looked on the phone.

Small pictures are not blown up

An icon dropped onto an A4 page is printed at its own size and centred, not stretched to fill the sheet. Enlarging it would add nothing but soft edges. Anything larger than the page is scaled down to fit, keeping its proportions exactly.

Why nothing is uploaded

Every other converter you will find asks you to upload the file. That is a reasonable way to build one and an unreasonable thing to do with a contract, a passport scan or a medical letter, which are exactly the documents people most often need as PDFs.

This page is served with a Content-Security-Policy that sets connect-srcto none. That is not a promise in a privacy policy; it is an instruction to your browser to refuse this page a network connection at all. If the code here tried to send your file somewhere, the browser would block the attempt and log it.

How to check it in a minute

Open your browser's developer tools, go to the network panel, and convert something. Nothing appears. Or switch your network off after the page has loaded and convert anyway: it works, because there is nothing on the other end to be missing. Both take less time than reading a policy that says the same thing.

What is kept on this device

A short list of recent conversions, holding a name, a few words of detail and a date. There is nowhere in that record for a file to go, which is what makes the line beside the list true rather than reassuring. Clearing it removes it. A finished PDF is held in memory until you download it and is gone when the page reloads.