Your files stay on your Mac.
Convertessa pulls the embedded text layer out of a PDF and writes it as UTF-8 plain text, one file per document, entirely on your Mac. It is the quick route from a paper, contract or manual to something you can grep, diff, word-count or pipe into a script. Point it at one file or a folder of hundreds and it works through the queue in a single pass.
A PDF is a page-description program, not a document in the sense a word processor means. Inside, each page is a list of drawing operators: select this font, move to these coordinates, show this run of glyphs. There is no paragraph object, no sentence, no reading order, only marks placed on a canvas in whatever sequence the producing application happened to emit them. Extracting text means reconstructing prose from that arrangement.
The extractor walks those content streams, maps each glyph back to a Unicode character through the font's ToUnicode table, and groups marks into lines by position. When the PDF came out of LaTeX, Word, InDesign or a browser's print engine, this produces clean, readable output. When the PDF is unusual, the seams show.
Two things deserve bluntness. Convertessa does not run OCR, so a scanned or photographed PDF whose pages are simply images with no text layer produces an empty or near-empty file; the tell is that you cannot select any text in Preview either. And plain text is lossless in the sense that nothing is compressed or degraded once the characters are written, which is why no quality control appears for this target. The losses are structural instead: fonts, bold and italic, images, links, annotations and page geometry have nowhere to go in a .txt file.
Drag in one document or a folder of them. Password-protected files fail at open, so decrypt those first; everything else joins the queue and keeps its filename, with the extension swapped.
Pick Plain Text as the output. No quality control appears here, because a text file is only characters and there is nothing to trade away. Each PDF produces its own matching text file.
Open one output before you trust the batch. Interleaved columns or runs of gibberish point at an awkward source layout or a font with no character map, rather than a failed conversion.
Images, audio, video, documents, archives, eBooks and fonts — all converted natively, right on your Mac.
Pick a format you have — see everything Convertessa can turn it into.
Convert hundreds of files at once with per-type group defaults and individual overrides.
First-class support for HEIC, AVIF and WebP — encode and decode, both ways. Read JPEG XL and a dozen more modern formats too.
Drop files anywhere on the window. Native, instant, and exactly what you’d expect on a Mac.
Dial in compression, resolution and bitrate. Keep originals pristine or shrink for sharing.
Never overwrite a file by accident. Convertessa appends safe suffixes automatically.
Every job is logged locally. Re-run a previous conversion or revisit recent outputs in a click.
Bundle your converted files straight into a single ZIP, TAR or 7z — packaged and ready to share the moment a job finishes.
Right-click any file in Finder and choose Convert. A Quick Action handles it on the spot — no need to open the app first.
Shrink Convertessa to a tidy mini window that tucks into a corner — just a drop zone and a format picker for quick, one-off conversions.
No cloud. No account. No upload. Every conversion happens entirely on-device, using the power already in your Mac. What you convert is nobody’s business but yours.
Script conversions, wire them into your build, or batch a folder from the terminal. The same engine, no GUI required.
Read the full CLI documentation →Shrink Convertessa down to a small, focused window — drop, convert, done, with your recent conversions one click away.
macOS 14+ · Apple Silicon & Intel
New formats, new features, and fixes — every release, on the record.
Almost always because the PDF has no text layer. Scans, photographed pages and faxes are images wrapped in a PDF, and there are no characters in the file to extract. Test it in Preview: if you cannot drag-select a word on the page, no extractor can read it either. You need recognition software to create a text layer first, then convert the resulting PDF.
Because columns are a visual illusion produced by glyph coordinates, not a structure stored in the file. Some producers emit a whole column before starting the next and extract perfectly; others alternate line by line and you get a zip-merge of both columns. If the paper is also published as HTML or as a LaTeX source, either will give you far cleaner text than the print PDF will.
No, there is no page-range option. Convertessa converts whole documents. In practice you convert once and then slice the result with tools already on your Mac:
convertessa report.pdf --to txt -o ~/Out && sed -n '1,120p' ~/Out/report.txt > intro.txt
That is usually faster than picking pages by hand anyway, because you can search the full text first and then cut at the line you actually care about.
Figures vanish entirely: they are images, and a text file cannot hold them. Tables survive only as their words. A row becomes a sequence of cell contents separated by spaces, with no reliable marker for where one column ends and the next begins, because the ruling lines were drawn as graphics. If you need the table back as a table, a spreadsheet or a Word target is the better destination.
Often not. AcroForm values live in the annotation layer rather than the page content stream, so an extractor reading the page can miss them completely while the values still show on screen. The reliable workaround is to flatten the form first: open it in Preview, print to PDF, and convert that copy. Flattening bakes the field values into the page content where extraction can see them.
The PDF's. Every line break you see in the output corresponds to a typeset line on the page, so a paragraph arrives as a stack of short lines wrapped at the original column width rather than as one flowing paragraph. Text editors that soft-wrap will look odd until you reflow. If you need reflowable prose, Markdown, DOCX or EPUB are better targets than plain text.
UTF-8. Curly quotes, em dashes, accented characters and typographic ligatures survive as their real Unicode codepoints, which means a word like find can arrive containing U+FB01 rather than the letters f and i. That trips naive search scripts. If a downstream tool insists on plain ASCII, run the output through iconv -f UTF-8 -t ASCII//TRANSLIT in.txt > out.txt before using it.