How to extract text from a PDF

Extracting text from a PDF is harder than it looks. PDF files are designed to preserve document appearance rather than facilitate text extraction. This post covers text extraction across multiple programming languages and compares tools for Python, JavaScript, Java, Swift, and Kotlin.

Why PDF text extraction is difficult

PDFs prioritize consistent rendering across devices, not text readability. They store low-level drawing commands rather than structured content, which makes extraction a heuristic problem.

How text is represented in a PDF

A PDF file doesn’t simply contain text as you’d be used to in a text file. What it does contain are commands on how to render the given text on the screen without whitespace characters or newlines. Here’s how PDFs represent text internally.

Content streams

Each page in a PDF has one or more content streams that tell the PDF viewer application how to render a page. A very simple one might look like this:

193.95 581.633 Td
(Hello) Tj
30.68 0 Td
(World!) Tj

These content streams can be represented differently while accomplishing the same goal, like this:

193.95 581.633 Td
<00290046004d004d00500001003800500053004d0045> Tj

Td instructs the PDF viewer where to draw the next string. Tj specifies which string to draw.

Extracting text from a content stream

The only way to extract text from a PDF is by looking at the rendering commands and making a heuristic attempt to reconstruct it. In the example above, we know we’re supposed to render Hello, reposition the text cursor, and then output World!.

You might have noticed there’s no whitespace in the first example above. Because the content stream only instructs the rendering engine what to draw on the screen, and because whitespace has no rendering command, we have to infer the spaces and newlines ourselves most of the time.

Doing this reliably across all the different PDF documents out there is difficult, and it’s not uncommon to encounter problems where tweaking the heuristic breaks one document but fixes another.

Extracting text with the Nutrient SDK

The Nutrient SDK offers APIs to retrieve text from a document across iOS, Android, and Web. All platforms use the same underlying heuristic to determine the layout of the text on the page and how to extract blocks out of it.

iOS

On iOS, you can use PSPDFTextParser to retrieve the text, text blocks, words, or glyphs from a page:

guard let textParser = documentProvider.textParserForPage(at: 0) else {
    // Handle failure.
    abort()
}
print("Text of page 0: \(textParser.text)")

for textBlock in textParser.textBlocks {
    print("TextBlock at \(textBlock.frame): \(textBlock.content)")
}

Android

On Android, there’s no dedicated text parser class; instead, you retrieve your page text using PdfDocument:

val pageText = document.getPageText(0)
print("Text of page 0: $pageText")

for (textRect in document.getPageTextRects(0, 0, pageText.length)) {
    val blockText = document.getPageText(0, textRect)
    print("TextBlock at $textRect: $blockText")
}

Web

Nutrient Web SDK can extract the text from a page using textLinesForPageIndex, but doesn’t currently support text block extraction:

const textLines = await instance.textLinesForPageIndex(0);
textLines.forEach((textLine) => console.log(textLine.contents));

instance.textLinesForPageIndex(0).then(function (textLines) {
  textLines.forEach(function (textLine) {
    console.log(textLine.contents);
  });
});

Python

Python has several extraction libraries. PyPDF is widely used and open source. PyMuPDF is faster and better at preserving layout. Our Python PDF text extraction tutorial covers both in depth:

from pypdf import PdfReader

reader = PdfReader("document.pdf")
for page in reader.pages:
    text = page.extract_text()
    print(text)

For scanned documents, you’ll need OCR. The Nutrient API handles both native text and scanned PDFs:

import os
import requests

with open("scanned.pdf", "rb") as f:
    response = requests.post(
        "https://api.nutrient.io/build",
        headers={"Authorization": f"Bearer {os.environ['NUTRIENT_API_KEY']}"},
        files={"file": f},
        data={
            "instructions": '{"parts": [{"file": "file"}], "actions": [{"type": "ocr"}], "output": {"type": "json-content"}}'
        },
        timeout=60,
    )
response.raise_for_status()
print(response.json())

Java

Java developers commonly use Apache PDFBox for text extraction. The example below uses the PDFBox 2.x API (PDDocument.load); on PDFBox 3.x, use Loader.loadPDF(new File(...)) instead:

import org.apache.pdfbox.pdmodel.PDDocument;
import org.apache.pdfbox.text.PDFTextStripper;

PDDocument document = PDDocument.load(new File("document.pdf"));
PDFTextStripper stripper = new PDFTextStripper();
String text = stripper.getText(document);
System.out.println(text);
document.close();

For more control over layout, PDFBox also offers PDFTextStripperByArea to extract text from specific regions of a page — useful when processing invoices or forms with known layouts.

Comparing PDF text extraction tools

Choose based on language, OCR requirements, and document complexity:

For a deeper comparison of Python options, see our PyMuPDF vs. Nutrient guide. For JavaScript, our complete guide to PDF.js covers the PDF.js approach in detail.

Handling scanned PDFs with OCR

Not all PDFs contain selectable text. Scanned documents store pages as images, so you need optical character recognition (OCR) to extract text from them. Common OCR approaches include Tesseract for open source workflows and the Nutrient API for production-grade accuracy. Our automated PDF OCR workflows guide walks through setting up batch processing for scanned documents. If you’re working with C#, see our guide on OCR for images and PDFs in C#.

Conclusion

PDF text extraction comes down to two things: how PDFs store text internally, and which tool fits your language and workflow. Whether you use Python, JavaScript, Java, or the Nutrient SDK on mobile, the core challenge is the same — PDFs store rendering commands, not structured text. The Nutrient SDK and Nutrient API handle both native text and scanned PDFs without requiring separate OCR tooling. This text extraction capability also powers our PDF text comparison feature for identifying differences between document versions. For a business perspective on document data extraction, see our tech leader’s guide to PDF data extraction.

FAQ

Related

Python text extraction: Parse PDFs with Python (PyPDF tutorial), extract text from PDFs with PyMuPDF, top 10 ways to generate PDFs in Python

JavaScript text extraction: Extract text from a PDF using JavaScript, complete guide to PDF.js, JavaScript PDF libraries compared

OCR and scanned PDFs: Use Tesseract OCR in Python, automated PDF OCR workflows, OCR images and PDFs in C#

Data extraction: PDF data extraction developer guide, tech leader’s guide to PDF data extraction

Document conversion: Convert PDF to JPG using Python, build an HTML5 PDF viewer