vs Python PDF Libraries

PDF Oxide compared with PyMuPDF (fitz), pypdfium2, pypdf, pdfplumber, pdfminer, and more. This page covers performance, feature coverage, licensing, and API differences to help you choose the right Python PDF library for text extraction.

Summary

	PDF Oxide	PyMuPDF	pypdfium2	pypdf	pdfplumber	pdfminer
Mean extraction time	0.8ms	4.6ms	4.1ms	12.1ms	23.2ms	16.8ms
Pass rate (3,830 PDFs)	100%	99.3%	99.2%	98.4%	98.8%	98.8%
License	MIT	AGPL-3.0	Apache-2.0	BSD-3	MIT	MIT
Language	Rust + PyO3	C (MuPDF)	C (PDFium)	Pure Python	Pure Python	Pure Python
Text extraction	Yes	Yes	Yes	Yes	Yes	Yes
Character positions	Yes	Yes	Yes	Partial	Yes	Yes
Image extraction	Yes	Yes	Yes	Yes	No	No
Form fields	Read + Write	Read + Write	Read only	Read + Write	Read only	No
PDF creation	Yes	Yes	No	Limited	No	No
PDF editing	Yes	Yes	No	Yes	No	No
Markdown output	Yes	No	No	No	No	No
HTML output	Yes	No	No	No	No	No
Encryption	Read + Write	Read + Write	Read only	Read + Write	No	No
PDF/A validation	Yes	No	No	No	No	No
Rendering	Yes	Yes	Yes	No	No	No
Search	Regex + spatial	Yes	Yes	No	No	No
Python versions	3.8–3.14	3.8–3.12	3.8+	3.6+	3.8+	3.6+
Install size	~5 MB wheel	~20 MB wheel	~3 MB wheel	~1 MB	~1 MB	~1 MB

Performance Comparison

Mean text extraction time per PDF, benchmarked on the full 3,830-PDF corpus — three independent, publicly available test suites that together cover every PDF specification version (1.0–2.0), encrypted files, malformed documents, CJK encodings, complex layouts, and security edge cases. See full corpus details for what each suite tests and why these results are reproducible.

Library	Mean	Relative	p99	Pass Rate
PDF Oxide	0.8ms	1×	9ms	100%
PyMuPDF	4.6ms	5.8×	28ms	99.3%
pypdfium2	4.1ms	5.1×	42ms	99.2%
pymupdf4llm	55.5ms	69×	280ms	99.1%
pdftext	7.3ms	9.1×	82ms	99.0%
pdfminer	16.8ms	21×	124ms	98.8%
pdfplumber	23.2ms	29×	189ms	98.8%
markitdown	108.8ms	136×	378ms	98.6%
pypdf	12.1ms	15.1×	97ms	98.4%

PDF Oxide achieves its speed through a native Rust core compiled to a Python extension module via PyO3. There is no subprocess overhead or C library bridging — the Rust code runs directly in the Python process.

Reliability

PDF Oxide processes 3,823 of 3,823 valid PDFs without failure — a 100% pass rate. The 7 non-passing files in the 3,830-file corpus are intentionally broken test fixtures (missing PDF header, fuzz-corrupted catalogs, invalid xref streams).

Library	Valid PDFs Passed	Pass Rate
PDF Oxide	3,823 / 3,823	100%
PyMuPDF	3,796 / 3,823	99.3%
pypdfium2	3,792 / 3,823	99.2%
pymupdf4llm	3,787 / 3,823	99.1%
pdftext	3,784 / 3,823	99.0%
pdfminer	3,777 / 3,823	98.8%
pdfplumber	3,777 / 3,823	98.8%
markitdown	3,771 / 3,823	98.6%
pypdf	3,762 / 3,823	98.4%

Text Quality

PDF Oxide achieves 99.5% text parity compared to PyMuPDF and pypdfium2 across the full corpus. Quality was measured by comparing extracted text output character-by-character. The remaining 0.5% difference is in whitespace normalization and ligature handling where PDF Oxide produces cleaner output.

License Comparison

Library	License	Commercial Use	Copyleft
PDF Oxide	MIT	Unrestricted	No
pypdfium2	Apache-2.0	Unrestricted	No
PyMuPDF	AGPL-3.0	Requires commercial license ($)	Yes
pypdf	BSD-3	Unrestricted	No
pdfplumber	MIT	Unrestricted	No
pdfminer	MIT	Unrestricted	No
pdftext	GPL-3.0	Requires open source	Yes

PyMuPDF uses MuPDF under the AGPL-3.0 license. If you distribute software that uses PyMuPDF, your software must also be released under AGPL-3.0 — or you must purchase a commercial license from Artifex. This applies to SaaS products, web applications, and any distributed binaries.

PDF Oxide is MIT-licensed with no restrictions. Use it in proprietary products, SaaS platforms, or closed-source applications without any licensing obligations.

Use Case	PDF Oxide (MIT)	PyMuPDF (AGPL)	pypdfium2 (Apache)	pypdf (BSD)	pdfplumber (MIT)	pdfminer (MIT)
Commercial product	Yes	Requires license	Yes	Yes	Yes	Yes
Closed source	Yes	No (unless licensed)	Yes	Yes	Yes	Yes
SaaS/cloud	Yes	Requires license	Yes	Yes	Yes	Yes
Internal tools	Yes	Yes	Yes	Yes	Yes	Yes

API Comparison

Text Extraction

PDF Oxide:

from pdf_oxide import PdfDocument

doc = PdfDocument("report.pdf")
text = doc.extract_text(0)
print(text)

PyMuPDF:

import fitz

doc = fitz.open("report.pdf")
page = doc[0]
text = page.get_text()
print(text)

pypdf:

from pypdf import PdfReader

reader = PdfReader("report.pdf")
page = reader.pages[0]
text = page.extract_text()
print(text)

pdfplumber:

import pdfplumber

with pdfplumber.open("report.pdf") as pdf:
    page = pdf.pages[0]
    text = page.extract_text()
    print(text)

pdfminer:

from pdfminer.high_level import extract_text

text = extract_text("report.pdf", page_numbers=[0])
print(text)

Character-Level Extraction

PDF Oxide:

from pdf_oxide import PdfDocument

doc = PdfDocument("report.pdf")
chars = doc.extract_chars(0)
for ch in chars:
    print(f"'{ch.char}' at ({ch.bbox[0]:.1f}, {ch.bbox[1]:.1f}) "
          f"size={ch.font_size:.1f}")

PyMuPDF:

import fitz

doc = fitz.open("report.pdf")
page = doc[0]
blocks = page.get_text("dict")["blocks"]
for block in blocks:
    if "lines" in block:
        for line in block["lines"]:
            for span in line["spans"]:
                print(f"'{span['text']}' size={span['size']:.1f}")

pdfplumber:

import pdfplumber

with pdfplumber.open("report.pdf") as pdf:
    page = pdf.pages[0]
    for char in page.chars:
        print(f"'{char['text']}' at ({char['x0']:.1f}, {char['top']:.1f}) "
              f"size={char['size']:.1f}")

pdfminer:

from pdfminer.high_level import extract_pages
from pdfminer.layout import LTChar

for page_layout in extract_pages("report.pdf"):
    for element in page_layout:
        if hasattr(element, '__iter__'):
            for text_line in element:
                if hasattr(text_line, '__iter__'):
                    for char in text_line:
                        if isinstance(char, LTChar):
                            print(f"'{char.get_text()}' at ({char.x0:.1f}, {char.y0:.1f}) "
                                  f"size={char.size:.1f}")

Image Extraction

PDF Oxide:

from pdf_oxide import PdfDocument

doc = PdfDocument("report.pdf")
images = doc.extract_image_bytes(0)
for i, img in enumerate(images):
    with open(f"image_{i}.{img['format']}", "wb") as f:
        f.write(img["data"])

PyMuPDF:

import fitz

doc = fitz.open("report.pdf")
page = doc[0]
for i, img in enumerate(page.get_images()):
    xref = img[0]
    base_image = doc.extract_image(xref)
    with open(f"image_{i}.{base_image['ext']}", "wb") as f:
        f.write(base_image["image"])

pypdf:

from pypdf import PdfReader

reader = PdfReader("report.pdf")
page = reader.pages[0]
for i, image in enumerate(page.images):
    with open(f"image_{i}.{image.name.split('.')[-1]}", "wb") as f:
        f.write(image.data)

PDF Creation

PDF Oxide:

from pdf_oxide import Pdf

pdf = Pdf.from_markdown("# Hello World\n\nThis is a PDF.")
pdf.save("output.pdf")

# Also supports HTML
pdf = Pdf.from_html("<h1>Hello</h1><p>World</p>")
pdf.save("output.pdf")

PyMuPDF:

import fitz

doc = fitz.open()
page = doc.new_page()
text_point = fitz.Point(72, 72)
page.insert_text(text_point, "Hello World", fontsize=24)
doc.save("output.pdf")

pypdf:

# pypdf can merge/modify PDFs but cannot create from scratch with text content.
# Use reportlab or fpdf2 for creation, then merge with pypdf.

Encrypted PDFs

PDF Oxide:

from pdf_oxide import PdfDocument

doc = PdfDocument("encrypted.pdf", password="password")
text = doc.extract_text(0)

PyMuPDF:

import fitz

doc = fitz.open("encrypted.pdf")
doc.authenticate("password")
page = doc[0]
text = page.get_text()

pypdf:

from pypdf import PdfReader

reader = PdfReader("encrypted.pdf")
reader.decrypt("password")
text = reader.pages[0].extract_text()

Markdown and HTML Output

PDF Oxide (unique feature):

from pdf_oxide import PdfDocument

doc = PdfDocument("paper.pdf")

# Convert to Markdown with heading detection
md = doc.to_markdown(0, detect_headings=True)
print(md)

# Convert to HTML
html = doc.to_html(0)
print(html)

No other Python PDF library provides built-in Markdown or HTML conversion.

Library Profiles

PDF Oxide

Strengths:

• Fastest text extraction in benchmarks due to Rust core — 5.8× faster than PyMuPDF
• 100% pass rate on 3,830-PDF corpus — highest reliability of any tested library
• Unified API for extraction, creation, and editing in a single library
• Built-in Markdown and HTML export with heading detection
• MIT licensed with no copyleft restrictions
• Native compliance validation (PDF/A, PDF/UA, PDF/X)
• Pre-built wheels for all major platforms and Python 3.8–3.14
• No system dependencies — the wheel includes everything

Limitations:

• Newer library with a smaller community
• Table extraction is basic compared to pdfplumber’s algorithms
• Rendering engine is less mature than MuPDF

PyMuPDF (fitz)

Strengths:

• Mature and battle-tested (backed by MuPDF, in development since 2005)
• Excellent rendering quality for complex PDFs
• Built-in OCR integration (Tesseract)
• Rich feature set: SVG export, page manipulation, table detection

Limitations:

• AGPL-3.0 license requires open-sourcing your application or purchasing a commercial license
• Large wheel size (~20 MB) due to bundled MuPDF
• No built-in Markdown export
• No compliance validation

pypdfium2

Strengths:

• Fast (backed by Google’s PDFium engine)
• Apache-2.0 license — permissive for commercial use
• Good rendering quality

Limitations:

• Limited text extraction API compared to PDF Oxide or PyMuPDF
• No PDF creation or editing
• No form field support beyond read-only

pypdf

Strengths:

• Pure Python — installs anywhere, no compiled dependencies
• Lightweight and well-maintained
• Good for PDF manipulation (merge, split, rotate, encrypt)
• Large community and extensive documentation

Limitations:

• 15× slower than PDF Oxide for text extraction
• Text extraction quality struggles with complex layouts
• No rendering, no Markdown/HTML export, no table extraction

pdfplumber

Strengths:

• Best table extraction of any Python PDF library
• Excellent character-level positioning data
• Visual debugging tools (annotated page images)
• MIT licensed

Limitations:

• Pure Python — 29× slower than PDF Oxide
• Read-only — no PDF creation or editing
• No encryption or rendering

pdfminer

Strengths:

• Detailed character and layout analysis
• Good CJK text support
• Foundation for pdfplumber and other tools
• MIT licensed

Limitations:

• 21× slower than PDF Oxide (pure Python, unoptimized)
• Read-only, no creation or editing
• Verbose API for common tasks
• Less actively maintained

When to Use Each

Use Case	Recommended Library
Fast text extraction	PDF Oxide
Commercial / proprietary product	PDF Oxide, pypdfium2, pypdf, pdfplumber, or pdfminer
PyMuPDF alternative (MIT licensed)	PDF Oxide
PDF creation from Markdown/HTML	PDF Oxide
Compliance validation (PDF/A, PDF/X)	PDF Oxide
Table extraction from invoices	pdfplumber
Visual debugging of extraction	pdfplumber
Existing MuPDF investment	PyMuPDF (if AGPL-compatible)
Minimal dependencies	pypdf (pure Python)
Detailed layout analysis	pdfminer
OCR for scanned documents	PyMuPDF

Installation

# PDF Oxide
pip install pdf_oxide

# PyMuPDF
pip install pymupdf

# pypdfium2
pip install pypdfium2

# pypdf
pip install pypdf

# pdfplumber
pip install pdfplumber

# pdfminer
pip install pdfminer.six

PDF Oxide ships pre-built wheels for Linux (x86_64, aarch64), macOS (x86_64, arm64), and Windows (x86_64). No compiler or system libraries required.

Related Pages

• Performance Benchmarks – full corpus benchmark results
• Getting Started with Python – installation and first extraction
• Python API Reference – complete Python API
• vs Rust PDF Libraries – Rust ecosystem comparison