Word to PDF Converter - Excel and PowerPoint

Convert Word to PDF online, along with Excel, PowerPoint, OpenDocument, RTF and plain text files. Upload one document or a hundred, get a PDF back for each - including archival PDF/A output. No Microsoft Office licence, no desktop install, no watermark stamped across the result.

The Actor is named for Word because that is what people search for, but it converts the whole Office family. Writer handles .docx, .doc, .odt, .rtf and .txt; Calc handles .xlsx, .xls, .ods and .csv; Impress handles .pptx, .ppt and .odp. It is the same code path each time - the file is read, the module that owns it is worked out from the bytes rather than the file extension, and LibreOffice exports the PDF.

• 📄 Converts Word, Excel and PowerPoint - twelve input formats, one PDF out
• 📁 Converts a whole folder in one run - up to 100 documents, each becoming its own PDF, all of them also packed into a single ZIP. One bad file does not stop the rest
• 🏛️ Archival PDF/A-1b, PDF/A-2b and PDF/A-3b - embeds every font and forbids anything whose appearance depends on the machine opening it, which is what archives and tax authorities ask for
• 🔖 Keeps headings as PDF bookmarks - a long report stays navigable in the reader's sidebar
• 🔤 Metric-compatible fonts are installed - Carlito for Calibri, Liberation for Arial and Times New Roman, so a document keeps its line breaks and its page count
• 🖼️ Image quality and downsampling controls - the practical way to get a photo-heavy deck under an email size limit
• 🔌 Runs from the API, a schedule or an integration - it is an Apify Actor, so Zapier, Make, n8n and MCP clients can call it

How to convert Word to PDF

1. Click Try for free and sign in to Apify.
2. Upload your .docx file in the Word file field, or paste a direct URL to it.
3. Leave PDF version on Standard PDF unless you need an archival file.
4. Click Start.
5. Download the PDF from the Storage tab, or from the download URL in the run's dataset record.

How to convert Excel to PDF

Identical steps with an .xlsx, .xls, .ods or .csv file. The result is what Calc would print, which is worth knowing: a spreadsheet with no print range set and a hundred columns of data becomes a great many pages. Set the print area, the page orientation and the scaling in Excel before converting, and the PDF will match.

How to convert PPT to PDF

Upload a .ppt, .pptx or .odp file. One slide becomes one page at the slide's own aspect ratio, so a 16:9 deck produces a 16:9 PDF rather than a letterboxed A4 one. Speaker notes are not exported.

How to convert DOCX to PDF

.docx is the ordinary modern Word format and the one this Actor is built around. .doc - the pre-2007 binary format - goes through the same path, and so do the macro-enabled .docm and the OpenDocument .odt. Nothing extra to configure: the Actor reads the document's content types part to see what it is, so a file with the wrong extension still reaches the right module.

How to convert PPTX to PDF

Same as .ppt above. .pptx generally converts with better fidelity than the old binary .ppt, because it carries explicit layout information rather than relying on a rendering engine's conventions.

How to convert RTF, TXT and CSV to PDF

These three have no container to inspect, so they are decided by their extension - which matters, because the same comma-separated bytes are a text document if you call them .txt and a spreadsheet if you call them .csv. CSV files are read as comma-separated UTF-8 with double-quote text delimiters, deliberately fixed rather than guessed.

How to convert a folder of documents to PDF in one run

Select every file in the Documents field rather than one, or pass an array of URLs to files from the API. Up to 100 documents per run. Each one becomes its own PDF in the run's storage, under its own filename, and all of them are also packed into a single documents.zip so there is one thing to download rather than fifty.

Mixing formats in one batch is fine and is the ordinary case: a folder of Word documents, spreadsheets and decks converts in a single run, because the module that opens each file is worked out per file.

One bad file does not cost you the batch. A password-protected document, a URL that returns a login page, a .docx that is actually a JPEG - each is recorded against its own row in the dataset with the reason, and the run carries on. The run only fails if nothing at all converted. The log ends with a count: converted 47, failed 3.

This is also the cheap way to use the Actor. See the cost section below: the container startup you pay for once is the same whether you convert one document or a hundred.

Input

Upload the files in the Documents field, or pass direct URLs to them.

Field	Description
files	The documents to convert. File uploads or URLs, up to 100 per run.
file	A single document. Hidden in the Console and kept for API callers and saved input configurations written before files existed. Ignored when files is filled in.
pdfVersion	default, 1.5, 1.6, 1.7, pdfa-1b, pdfa-2b or pdfa-3b.
outputFilename	Defaults to the input filename with a .pdf extension. In a batch it names the ZIP instead, and each PDF keeps its own filename.
exportBookmarks	true by default. Turns headings and bookmarks into PDF outline entries.
imageQuality	1 to 100, default 90. JPEG quality for images inside the document.
maxImageResolution	0 (do not reduce), 75, 150, 300, 600 or 1200 DPI.
fileBase64, filename	A single file as base64, for API callers who prefer JSON to a URL.
verboseLog	Include the full LibreOffice output in the Actor log.

Which PDF version should I choose?

Standard PDF for anything you are going to email, print or upload. It is the smallest of the options and every reader written this century opens it. LibreOffice currently writes PDF 1.6 for this setting; the exact version is reported in the run's dataset record.

PDF/A when the file has to survive. The profile embeds every font, forbids transparency features whose rendering has drifted between readers, and records its own conformance level inside the file. Archives, courts, patent offices and tax authorities generally ask for PDF/A-1b or PDF/A-2b; PDF/A-3b is the same thing but permits arbitrary file attachments, which is how e-invoicing formats such as ZUGFeRD and Factur-X ship their XML inside the PDF. Archival files are larger, sometimes several times larger, because the fonts travel with them.

Output

Every converted PDF is written to the run's key-value store under its own filename, so it is downloadable from the Storage tab. Names are sanitised to the character set store keys allow, and collisions are de-duplicated rather than silently overwritten - which matters more here than in most converters, since every output is a .pdf, so a notes.docx and a notes.xlsx in the same batch become notes.pdf and notes-2.pdf.

OUTPUT is the record the run-sync endpoint returns in its HTTP response body, and it holds whatever a caller most likely wants:

Run	OUTPUT
One document	The PDF itself, so run-sync returns a PDF.
Several documents	A ZIP of every PDF, also stored as documents.zip.
A batch whose PDFs total more than 192 MB	A JSON manifest of the keys, sizes and download URLs. The archive has to be held in memory to be uploaded, so past that size it is skipped rather than risking the run; the individual PDFs are all still there.

One dataset record per input file, carrying ok, the source format, the module that converted it, the PDF version, file sizes and download URLs - and, for anything that failed, error with the reason in plain language. Records are pushed as each file finishes, so a run that dies at file 80 still has 79 results.

How much does it cost to convert Word to PDF?

Job	Run time	Cost
One Word letter	11.8 s	$0.0015
10 letters in one run	16.7 s	$0.0026, or $0.0003 each

Measured 2026-08-26 on the Free plan at the default 2048 MB, read back from the run API.

A document costs a fraction of a cent, and Apify's free plan includes $5 of usage a month.

Batch, and the cost of a folder is close to the cost of one document. That matters more here than anywhere else in the family, because LibreOffice is large: the image is roughly 1.4 GB even with only Writer, Calc and Impress, and starting it costs more wall-clock time than converting an ordinary document does. Ten documents in one run took 16.7 s against 11.8 s for one - five extra seconds for nine extra documents. As ten separate runs it would have been about $0.015 rather than $0.0026. Put the whole folder in the Documents field.

The user profile is prebuilt into the image for the same reason. LibreOffice writes one the first time it starts, and doing that inside the run would add several seconds to every conversion.

Memory is set to 2048 MB rather than the usual 1024 MB because LibreOffice needs a working set of its own before it opens anything. Apify allocates one CPU core per 4096 MB. A batch does not need more than a single file, because documents are converted strictly one at a time and each scratch directory is deleted as soon as its PDF is uploaded.

Limits on a single run

Limit	Value
Documents per run	100
Total input size per run	512 MB
One document	100 MB
Download timeout per file	2 minutes
Conversion timeout per document	10 minutes
ZIP size, past which a manifest is written instead	192 MB

A run over any of these says which number it exceeded. The file count is checked before anything is downloaded, so a folder of 5,000 is refused immediately rather than at file 100.

How to convert Word to PDF with the API

run-sync returns the PDF in the response body:

curl -X POST \
  "https://api.apify.com/v2/acts/marcel-rbro~word-to-pdf/run-sync?token=<TOKEN>" \
  -H 'Content-Type: application/json' \
  -d '{"file": "https://example.com/report.docx"}' \
  --output report.pdf

A folder is an array, and OUTPUT becomes a ZIP:

curl -X POST \
  "https://api.apify.com/v2/acts/marcel-rbro~word-to-pdf/run-sync?token=<TOKEN>" \
  -H 'Content-Type: application/json' \
  -d '{"files": ["https://example.com/a.docx", "https://example.com/b.xlsx", "https://example.com/c.pptx"]}' \
  --output documents.zip

Asking for an archival PDF is one more field:

curl -X POST \
  "https://api.apify.com/v2/acts/marcel-rbro~word-to-pdf/run-sync?token=<TOKEN>" \
  -H 'Content-Type: application/json' \
  -d '{"file": "https://example.com/report.docx", "pdfVersion": "pdfa-2b"}' \
  --output report-archival.pdf

If you hold the bytes rather than a URL, pass them as base64 in fileBase64. The Apify API requires Actor input to be application/json and rejects anything else with HTTP 400 before the run starts, so posting the document as a raw request body is not an option:

curl -X POST \
  "https://api.apify.com/v2/acts/marcel-rbro~word-to-pdf/run-sync?token=<TOKEN>" \
  -H 'Content-Type: application/json' \
  -d "{\"fileBase64\": \"$(base64 < report.docx)\", \"filename\": \"report.docx\"}" \
  --output report.pdf

Base64 adds about a third to the payload, so prefer a URL for large files. Pass filename with it: for .txt and .csv the extension is the only thing that says which module should open the file.

run-sync waits up to 300 seconds, and a batch of any size will exceed that long before it exceeds the run's own timeout. A run converting 100 documents keeps going for as long as it needs; the run-sync connection does not. So for anything past a handful of files, start the run with the asynchronous Run Actor endpoint, poll it, and read the results from the dataset and the key-value store when it finishes. run-sync is for the one-file case and for small batches.

There are client libraries for JavaScript and Python, and the same run can be triggered on a schedule or from Zapier, Make, n8n, Slack or an MCP client.

Limitations

• Fonts are substituted, not invented. The image carries Liberation, Carlito, Caladea, DejaVu and Noto, which cover Arial, Times New Roman, Courier New, Calibri, Cambria and a wide range of scripts including Chinese, Japanese and Korean. Those substitutes are metric-compatible, meaning the glyph widths match, so line breaks and page counts hold. A document set in a font that is none of those - a licensed corporate typeface, a font you bought - will be rendered in something else, and the shapes will differ even where the spacing does not. There is no way around this short of embedding the font in the source document, which OOXML supports and LibreOffice honours.
• Macros are not run. .docm, .xlsm and .pptm convert, but their macros do not execute, so anything a macro would have generated is absent. This is a security property rather than an oversight.
• Password-protected files fail. The Actor detects the encryption wrapper and stops with an explanation rather than waiting for a prompt nobody can answer. Remove the protection in Office first.
• Base and Math are not installed. The image carries Writer, Calc and Impress only, to keep an already large download from being a much larger one. An embedded Math formula object or a Base report will not render.
• Spreadsheets convert as they would print. No print range and no page scaling means a very long PDF. Set them in the source file.
• Speaker notes are not exported from presentations, and neither are slide transitions or animations, which a page cannot represent.
• CSV parsing is fixed, not guessed. Comma separator, double-quote text delimiter, UTF-8. A semicolon-separated export from a European Excel will land in a single column. Convert it to .xlsx first, or re-save it with commas.
• PDF/A conformance is claimed, not validated. LibreOffice writes the conformance metadata and follows the profile; this Actor does not run veraPDF over the result. If you need a certificate of validity, validate the output yourself.
• 100 documents and 512 MB of input per run. Past either, split the batch; the error says which limit was hit. A single document over 100 MB is refused on its own, because LibreOffice loads a file entirely into memory before it can lay it out.
• Documents are converted one at a time, deliberately. The image ships a prebuilt LibreOffice user profile so that runs do not pay for creating one, and two soffice processes must never share a profile - the second hands its command line to the first and exits reporting success, having converted nothing. Sequential reuse is safe; parallelising it is not. A batch is therefore no faster per document than a single-file run, it just stops paying for startup repeatedly.

Other file converters

Actor	Converts
Word to PDF converter (this one)	DOCX, DOC, ODT, RTF, TXT, XLSX, XLS, ODS, CSV, PPTX, PPT, ODP → PDF
HEIC to JPG converter	HEIC, HEIF, AVIF → JPEG, PNG, WebP, AVIF, TIFF
MOBI to EPUB converter	MOBI, PRC, AZW, AZW3 → EPUB
JPG to PDF converter	JPG, PNG → PDF
PDF to JPG converter	PDF → JPG, PNG
PNG to JPG converter	PNG → JPG, WebP
EPUB to PDF converter	EPUB → PDF

FAQ

Is this Word to PDF converter free?

Apify's free plan includes $5 of platform usage per month and needs no card. A document conversion costs a fraction of that; the exact figure is being measured and will appear in the cost section above.

Will the formatting survive?

Mostly, and the exceptions are predictable. Text, headings, tables, lists, headers and footers, page numbering and images all come through. What moves is anything that depended on a font the container does not have, and anything Word renders through a feature LibreOffice implements differently - complex text boxes anchored to floating shapes are the usual culprit, along with SmartArt and Word's newer chart types.

The font question is the one that decides most outcomes, which is why the image carries metric-compatible substitutes rather than leaving it to chance. Calibri becomes Carlito, Arial becomes Liberation Sans, Times New Roman becomes Liberation Serif, Cambria becomes Caladea. Those pairs share glyph widths, so a document keeps its line breaks and its page count even though the letterforms are not identical. Anything outside that set is substituted by whatever fontconfig thinks is closest, and the layout can reflow.

If exact fidelity matters more than anything else, embed the fonts in the source document before converting. Word does this from File → Options → Save → Embed fonts in the file, and LibreOffice reads them.

What is PDF/A, and do I need it?

PDF/A is the ISO profile for long-term archiving. It requires every font to be embedded, forbids external dependencies and features whose rendering has drifted between readers, and records its own conformance level in the file's metadata, so a reader twenty years from now can tell what it is looking at and render it the same way you saw it.

You need it when somebody asks for it: national archives, courts, patent offices, tax authorities and academic repositories generally do. PDF/A-1b is the strictest and most widely demanded, PDF/A-2b is the sensible modern default, and PDF/A-3b additionally allows arbitrary attachments, which is how e-invoicing standards such as ZUGFeRD and Factur-X carry their XML inside the PDF. For an ordinary document you are emailing to somebody, you do not need it, and the file will be smaller without it.

Can I convert Excel to PDF without the columns being cut off?

Set the print area, the orientation and the fit-to-width scaling in the source spreadsheet. The conversion produces what Calc would print, so it inherits whatever page setup the workbook carries. A workbook with none set prints at its natural width, which for a wide sheet means columns spilling onto separate pages.

Can I convert a PDF back to Word here?

No. This Actor goes one way. PDF to Word is a genuinely different and much harder problem - a PDF records where glyphs sit on a page, not what a paragraph is - and it deserves its own tool rather than a checkbox on this one.

Can I batch convert a folder of documents?

Yes, and it is the cheapest way to use it. Select every file in the Documents field, or pass an array of URLs to files from the API - up to 100 documents per run. Each becomes its own PDF, and all of them are also packed into a single ZIP so there is one download rather than fifty.

The saving is not marginal. This Actor's image is around 1.4 GB, and pulling and starting it takes longer than converting an ordinary document. A folder converted as one run pays that once instead of once per file.

A batch of any real size will outlast the 300-second run-sync window, so start it with the asynchronous endpoint and read the dataset when it finishes.

What happens if one document in the batch is broken?

That file fails and the rest carry on. Its dataset record gets ok: false and an error explaining what was wrong - password-protected, damaged, not actually an Office file, a URL that 404ed. The run only fails if nothing converted at all, and the log ends with a count of how many succeeded and how many did not.

Can an AI agent call this converter?

Yes. Every Apify Actor is exposed through the Apify MCP server, so Claude, Cursor and other MCP clients can run this conversion as a tool. The fileBase64 field exists partly for that case, since an agent holding bytes rather than a public URL is the common situation.

The conversion failed. What now?

Turn on verboseLog and re-run: the Actor log will carry LibreOffice's own output. The common causes are a password-protected document, a URL that returns an HTML login page rather than the file, and a document whose extension does not match its contents - the Actor reports what it actually detected, which usually explains the last of those on its own.

Development

npm install
npm test      # format inspector, batch bookkeeping and ZIP writer; no LibreOffice needed
apify run     # needs soffice on your PATH

Conversion runs LibreOffice's soffice in headless mode inside a Debian container. npm test covers the three pure modules - the format inspector (the zip, OpenDocument and OLE2 parsing that decides which LibreOffice module opens a file), the batch bookkeeping (key collisions and the run's budgets) and the ZIP writer - and deliberately needs no LibreOffice at all, so it runs in a second on any machine. The conversion itself is exercised by the Docker build, which converts a Writer document and a Calc document for real and fails the build if no PDF comes out.

src/zip.js is copied from the pdf-to-jpg Actor in this family rather than depended on: these Actors have exactly one runtime dependency, apify, and PDFs are already compressed, so every archive entry is stored rather than deflated.

This Actor's code is MIT licensed. LibreOffice is MPL-2.0 with LGPL-3 portions and runs as a separate process; the bundled fonts are under the SIL Open Font License, Apache-2.0 and the Bitstream Vera licence - see THIRD_PARTY_NOTICES.md.