How to extract text from PDFs server-side with PDF.js in Node.js

PDF.js isn’t just for browsers. You can use pdfjs-dist in Node.js to extract text from PDFs without any rendering. This is useful for search indexing and content analysis.

Install

Install the pdfjs-dist package:

npm install pdfjs-dist

Basic text extraction

The following example loads a PDF buffer, walks each page, and returns an array of per-page text strings:

import { getDocument, PDFDocumentProxy } from "pdfjs-dist";
import path from "path";
import { createRequire } from "node:module";

// Resolve paths for CMap and font data from the pdfjs-dist package.
const require = createRequire(import.meta.url);
const pdfjsDistPath = require.resolve("pdfjs-dist");
const fontDataPath = path.join(path.resolve(pdfjsDistPath, "../../standard_fonts"), "/");
const cmapPath = path.join(path.resolve(pdfjsDistPath, "../../cmaps"), "/");

async function extractText(pdfBuffer) {
  const uint8Array = new Uint8Array(pdfBuffer);

  const pdf = await getDocument({
    data: uint8Array,
    standardFontDataUrl: fontDataPath,
    cMapUrl: cmapPath,
  }).promise;

  const pageTexts = [];
  for (let i = 1; i <= pdf.numPages; i++) {
    const page = await pdf.getPage(i);
    const textContent = await page.getTextContent();

    const pageText = textContent.items.reduce((text, item) => {
      if ("str" in item && "hasEOL" in item) {
        return text + item.str + (item.hasEOL ? "\n" : "");
      }
      return text;
    }, "");

    pageTexts.push(pageText);
  }

  return pageTexts;
}

Understanding TextContent items

page.getTextContent() returns an object with an items array. Each item has:

{
  str: "Hello world",   // The text string.
  hasEOL: true,         // Whether it's followed by a line break.
  dir: "ltr",           // Text direction.
  width: 120.5,         // Width in PDF points.
  height: 12,           // Height in PDF points.
  transform: [12, 0, 0, 12, 72, 700], // Position/transform matrix.
}

Filter for items with str and hasEOL — some items are positioning markers without text.

Parallel page processing

For large PDFs, extract pages in parallel. The snippet below uses the pdf document from the basic example above:

const textPromises = [];
for (let i = 1; i <= pdf.numPages; i++) {
  textPromises.push(
    pdf.getPage(i)
      .then((page) => page.getTextContent())
      .then((textContent) =>
        textContent.items.reduce((text, item) => {
          if ("str" in item && "hasEOL" in item) {
            return text + item.str + (item.hasEOL ? "\n" : "");
          }
          return text;
        }, ""),
      ),
  );
}
const allTexts = await Promise.all(textPromises);

Configuration notes

PDF.js requires three adjustments when running in Node.js that differ from a browser setup.

CMap and font paths

In Node.js, you need to resolve these paths from the pdfjs-dist package itself:

const pdfjsDistPath = require.resolve("pdfjs-dist");
const fontDataPath = path.join(path.resolve(pdfjsDistPath, "../../standard_fonts"), "/");
const cmapPath = path.join(path.resolve(pdfjsDistPath, "../../cmaps"), "/");

These are directory paths to the bundled CMap and font files — required for proper text extraction from Chinese, Japanese, and Korean (CJK) and non-standard fonts.

No worker needed

In Node.js, you don’t need to configure GlobalWorkerOptions.workerSrc. PDF.js falls back to running in the main thread, which is fine for server-side batch processing.

Polyfill

If your Node.js version doesn’t support Promise.withResolvers, add a polyfill:

import "core-js/proposals/promise-with-resolvers";

Key points

• pdfjs-dist works in Node.js without a browser or canvas.
• getDocument({ data }) accepts Uint8Array from a Buffer.
• page.getTextContent() gives you structured text with position metadata.
• Filter text items for "str" in item to skip non-text markers.
• No worker configuration is needed for server-side use.
• Always configure CMap and standard font paths for proper text extraction.
• Text extraction is the foundation for search indexing and content analysis.

FAQ

How Nutrient Web SDK handles this

No Node.js PDF.js setup, no CMap path resolution, and no polyfills are required. Nutrient Web SDK extracts text from any page with a single API call:

// Extract text from a page — one line.
const textLines = await instance.textLinesForPageIndex(0);
console.log(textLines.map((line) => line.contents).join("\n"));

For server-side processing, Nutrient Document Engine provides text extraction, optical character recognition (OCR), and content analysis through a REST API, with no extra binary dependencies on the host.

See Nutrient Web SDK for an alternative to wiring up PDF.js in Node. The text extraction and OCR guides cover the API in depth. Follow the migration guide to switch, or talk to Sales about your requirements.