Chuyển Đổi PDF Sang Word DOCX Bằng Node.js

Converting PDF documents into editable Microsoft Word (.docx) files is a common requirement for legal portals, contract review platforms, and document management systems. Because PDF is primarily a display format (specifying absolute glyph coordinates) rather than a semantic structure, conversion requires structural reconstruction.

Conversion Approaches in Node.js

1. LibreOffice Headless (Self-Hosted): High fidelity, converts complex tables and layouts via command line.
2. Cloud Converter APIs (ConvertAPI, CloudConvert): Zero maintenance, handles OCR and rasterized text.
3. pdf2docx (Python bridge via Node child_process): High quality for vector documents and tables.

Using Headless LibreOffice with Node.js

Running LibreOffice in headless mode inside a Docker container provides free, reliable document conversion:

import { exec } from 'child_process';
import util from 'util';
import path from 'path';
import fs from 'fs';

const execPromise = util.promisify(exec);

export async function convertPdfToDocx(inputPdfPath: string, outputDir: string): Promise<string> {
  if (!fs.existsSync(inputPdfPath)) {
    throw new Error('Input PDF does not exist.');
  }

  // soffice is the executable name for LibreOffice
  const cmd = `soffice --headless --convert-to docx "${inputPdfPath}" --outdir "${outputDir}"`;

  console.log('Executing conversion:', cmd);
  const { stdout, stderr } = await execPromise(cmd);

  if (stderr && !stderr.includes('Warning')) {
    console.warn('LibreOffice warnings:', stderr);
  }

  const baseName = path.basename(inputPdfPath, path.extname(inputPdfPath));
  const expectedOutputPath = path.join(outputDir, `${baseName}.docx`);

  if (!fs.existsSync(expectedOutputPath)) {
    throw new Error('Conversion completed but output DOCX file was not generated.');
  }

  return expectedOutputPath;
}

Key Conversion Challenges

• Embedded fonts: Missing system fonts can cause text overlapping and reflow issues. Install standard Microsoft Core Fonts (msttcorefonts) in your Linux/Docker host.
• Scanned pages: PDFs containing only raster images require an Optical Character Recognition (OCR) step using Tesseract before conversion.