Convert Word to PDF without comments or tracked changes

Our DOCX to PDF prints a reviewed Word file as if every tracked change had been accepted, and leaves the comments out, unless you set IncludeMarkup=true. We converted a contract draft with insertions, deletions, a moved paragraph, two comments and a hidden sentence, then searched the PDF for each of them. The insertions print as ordinary text; none of the deleted text, the comments, the reviewers' names on them or the hidden sentence was in the file. What did come through from the Word file is in the PDF's document properties: the author's name, the title and the creation date. A second request removes the name.

# 1. Word to PDF. Without IncludeMarkup, tracked changes print as accepted and comments are left out.
curl -sS -X POST https://api.pdfconversionapi.com/v2/convert/docx/to/pdf \
  -H "Authorization: Bearer $PDFCONVERSIONAPI_KEY" \
  -F File=@supply-agreement-draft.docx \
  -o step1.json
jq -r '.Files[0].FileData' step1.json | base64 --decode > draft.pdf

# 2. Clear the Author property the PDF took from the Word file.
curl -sS -X POST https://api.pdfconversionapi.com/v2/convert/pdf/to/pdf \
  -H "Authorization: Bearer $PDFCONVERSIONAPI_KEY" \
  -F File=@draft.pdf \
  -F "PdfAuthor=' '" \
  -F FileName=supply-agreement \
  -o step2.json
jq -r '.Files[0].FileData' step2.json | base64 --decode > supply-agreement.pdf

What the PDF contains without IncludeMarkup

"Not in the file" means a phrase from that item was in none of these: the PDF's page text, its annotations, its document properties and XMP metadata, or the strings in its objects and decompressed streams. The PDF has no annotations at all, so the comments aren't hidden as PDF notes either.

The PDF's creation date also comes from the Word file: our draft was created on October 20, and the PDF's creation and modification dates both say October 20.

Remove the author's name

IncludeMetadata=false, the option for leaving the Word file's properties out, isn't supported on DOCX to PDF yet. The API refuses it rather than ignore it:

{"Code":400,"Message":"IncludeMetadata is not supported yet for DOCX → PDF"}

Instead, send the PDF to pdf/to/pdf with PdfAuthor=' ' (a space in single quotes), as in step 2 above. The Author property is removed, and "Reviewer A" no longer appears anywhere in the file. The page text, bookmarks and fonts stayed the same, and the page rendered pixel for pixel the same at 72 DPI. The modification date in the document information becomes the time of the second request; the copy in the XMP metadata still says October 20. The title, "Supply agreement, draft 3", stays; PdfTitle changes or removes it the same way (see PDF to PDF).

To check a PDF before you send it, ask for its metadata:

curl -sS -X POST https://api.pdfconversionapi.com/v2/convert/pdf/to/meta \
  -H "Authorization: Bearer $PDFCONVERSIONAPI_KEY" \
  -F File=@supply-agreement.pdf \
  -o meta.json

jq -r '.Files[0].FileData' meta.json | base64 --decode \
  | jq -c '{author, title, annotations: (.annotations | length), attachments: (.attachments | length)}'

{"author":null,"title":"Supply agreement, draft 3","annotations":0,"attachments":0}

Print the comments and changes: IncludeMarkup=true

curl -sS -X POST https://api.pdfconversionapi.com/v2/convert/docx/to/pdf \
  -H "Authorization: Bearer $PDFCONVERSIONAPI_KEY" \
  -F File=@supply-agreement-draft.docx \
  -F IncludeMarkup=true \
  -o response.json

jq -r '.Files[0].FileData' response.json | base64 --decode > supply-agreement-draft-markup.pdf

• Insertions are underlined and deletions struck through, in one color per reviewer, with a bar in the left margin beside each changed line.
• The moved paragraph shows twice, struck through where it was and underlined where it went, in a third color.
• Each comment sits in a column on the right with the reviewer's name, joined to the text it is about by a dotted line.
• To make room for that column, the page's content is scaled to 74% of its size on the same Letter page.
• The tracked bold change isn't shown as a change: "within 30 days" prints bold and black, with no note. Its line has a change bar, but that bar is for "twelve" becoming "eighteen" on the same line, so this file can't tell whether a formatting change gets a bar of its own.
• The hidden sentence is still left out.

Text extracted from this PDF runs the old and new values together: "within 3045 days", "twelveeighteen months". If you index markup PDFs for search, index the version without markup as well.

The same two steps in Python and Node.js

Each version writes supply-agreement.pdf without comments, tracked changes or the author's name, and prints its size: 31,577 bytes in our run, give or take a byte between runs, because the file records when it was saved.

# Needs the requests package: pip install requests
import base64
import os

import requests

API = "https://api.pdfconversionapi.com/v2/convert"
HEADERS = {"Authorization": f"Bearer {os.environ['PDFCONVERSIONAPI_KEY']}"}


def convert(route, name, data, options):
    response = requests.post(f"{API}/{route}", headers=HEADERS, files={"File": (name, data)}, data=options, timeout=150)
    result = response.json()
    if not response.ok:
        raise SystemExit(f"{route}: {result['Code']} {result['Message']}")
    return base64.b64decode(result["Files"][0]["FileData"])


def clean_pdf(path):
    """Word to PDF without comments or tracked changes, then without the Word file's author."""
    with open(path, "rb") as f:
        pdf = convert("docx/to/pdf", os.path.basename(path), f.read(), {})
    return convert("pdf/to/pdf", "draft.pdf", pdf, {"PdfAuthor": "' '"})


pdf = clean_pdf("supply-agreement-draft.docx")
with open("supply-agreement.pdf", "wb") as out:
    out.write(pdf)
print("supply-agreement.pdf", len(pdf), "bytes")

// Needs Node.js 20 or later. Save it as clean.mjs and run: node clean.mjs
import { readFile, writeFile } from "node:fs/promises";
import { basename } from "node:path";

const API = "https://api.pdfconversionapi.com/v2/convert";

async function convert(route, name, data, options = {}) {
  const form = new FormData();
  form.append("File", new Blob([data]), name);
  for (const [key, value] of Object.entries(options)) form.append(key, value);
  const response = await fetch(`${API}/${route}`, {
    method: "POST",
    headers: { Authorization: `Bearer ${process.env.PDFCONVERSIONAPI_KEY}` },
    body: form,
  });
  const result = await response.json();
  if (!response.ok) throw new Error(`${route}: ${result.Code} ${result.Message}`);
  return Buffer.from(result.Files[0].FileData, "base64");
}

/** Word to PDF without comments or tracked changes, then without the Word file's author. */
async function cleanPdf(path) {
  const pdf = await convert("docx/to/pdf", basename(path), await readFile(path));
  return convert("pdf/to/pdf", "draft.pdf", pdf, { PdfAuthor: "' '" });
}

const pdf = await cleanPdf("supply-agreement-draft.docx");
await writeFile("supply-agreement.pdf", pdf);
console.log("supply-agreement.pdf", pdf.length, "bytes");

What you can't do yet

• Print the original wording, as if every change had been rejected. There is no option for it.
• Print comments without tracked changes, or the reverse. IncludeMarkup is one switch for both.
• Keep comments as PDF notes that a reader can open and reply to. With IncludeMarkup=true they are drawn on the page; the PDF has no annotations.
• Accept or reject the changes in the Word file. docx/to/docx returned the draft with every tracked change and both comments still in it. Converting to PDF never touches your original .docx: if you send the Word file along with the PDF, the comments, deleted text and hidden sentence go with it.

One layout difference we found: the paragraph removed by the tracked move, just before the heading "2. Price and payment", left 8 points (about 3 mm) of extra space above that heading, compared with a copy of the file without that paragraph. A copy with the same paragraph as a tracked deletion instead left the same 8 points. The deleted paragraph between two ordinary paragraphs in section 2 left no gap we could measure. The text is unaffected.

How we tested