OCR in Python: Extract Text from Images and Scanned PDFs

Optical Character Recognition (OCR) converts images of text into machine-readable text. You need it when:

• You have scanned PDFs that are just images wrapped in a PDF container
• You need to extract text from photos of documents, receipts, or whiteboards
• You're building a document processing pipeline that handles both digital and scanned files
• You want to make scanned documents searchable

The two main approaches in 2026 are:

• 1. Tesseract OCR — free, open-source, runs locally, good for clean documents
• 2. AI-based OCR (GPT-4o Vision, Google Cloud Vision) — better accuracy on messy handwriting, costs per request

Tesseract is the most widely used open-source OCR engine. Install it first:

Ubuntu/Debian:

sudo apt update
sudo apt install tesseract-ocr tesseract-ocr-eng

macOS:

brew install tesseract

Windows:

Download the installer from the official UB-Mannheim repository on GitHub.

Verify the installation:

tesseract --version
# tesseract 5.x.x

Then install the Python wrapper:

pip install pytesseract Pillow pdf2image

The simplest use case — extract text from a single image:

import pytesseract
from PIL import Image

# Open image
img = Image.open("document.png")

# Extract text
text = pytesseract.image_to_string(img)
print(text)

For better accuracy, preprocess the image first:

import pytesseract
from PIL import Image, ImageFilter, ImageEnhance

def preprocess_for_ocr(image_path):
    img = Image.open(image_path)
    
    # Convert to grayscale
    img = img.convert("L")
    
    # Increase contrast
    enhancer = ImageEnhance.Contrast(img)
    img = enhancer.enhance(2.0)
    
    # Sharpen
    img = img.filter(ImageFilter.SHARPEN)
    
    # Binarize (convert to pure black and white)
    img = img.point(lambda x: 0 if x < 128 else 255)
    
    return img

img = preprocess_for_ocr("receipt.jpg")
text = pytesseract.image_to_string(img)
print(text)

Preprocessing typically improves accuracy by 20-40% on photos and scanned documents.

Scanned PDFs are trickier — you need to convert each page to an image first, then run OCR:

from pdf2image import convert_from_path
import pytesseract

def ocr_pdf(pdf_path):
    # Convert PDF pages to images (300 DPI for good quality)
    images = convert_from_path(pdf_path, dpi=300)
    
    full_text = []
    for i, page_image in enumerate(images):
        text = pytesseract.image_to_string(page_image)
        full_text.append(f"--- Page {i + 1} ---\n{text}")
    
    return "\n\n".join(full_text)

result = ocr_pdf("scanned-contract.pdf")
print(result)

Important: pdf2image requires poppler-utils to be installed:

# Ubuntu
sudo apt install poppler-utils

# macOS
brew install poppler

Tesseract supports 100+ languages. Install the language packs you need:

# Install German and French
sudo apt install tesseract-ocr-deu tesseract-ocr-fra

# List available languages
tesseract --list-langs

Then specify the language in your code:

# Single language
text = pytesseract.image_to_string(img, lang="deu")

# Multiple languages (if document mixes languages)
text = pytesseract.image_to_string(img, lang="eng+deu+fra")

For best results with non-Latin scripts (Arabic, Chinese, Japanese), use the _best trained data files instead of the default _fast ones.

Tesseract struggles with:

• Handwritten text
• Low-quality photos with poor lighting
• Complex layouts (tables, multi-column documents)
• Curved or distorted text

For these cases, AI vision models are significantly better. Here's how to use GPT-4o-mini for OCR:

from openai import OpenAI
import base64

client = OpenAI()

def ai_ocr(image_path):
    with open(image_path, "rb") as f:
        b64 = base64.b64encode(f.read()).decode()
    
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{
            "role": "user",
            "content": [
                {"type": "image_url", "image_url": {"url": f"data:image/png;base64,{b64}"}},
                {"type": "text", "text": "Extract all text from this image. Preserve formatting."}
            ]
        }],
        max_tokens=4000
    )
    return response.choices[0].message.content

text = ai_ocr("handwritten-notes.jpg")
print(text)

Cost: ~$0.001-0.005 per image with gpt-4o-mini. For free OCR without any setup, try Reformat's OCR tool — it handles images and scanned PDFs with a single upload.

Use Tesseract when:

• Documents are cleanly printed
• You need offline/local processing
• Budget is zero

Use AI OCR when:

• Documents have handwriting, tables, or complex layouts
• Accuracy is critical
• You're processing small volumes

For a quick one-off extraction without installing anything, Reformat's Image to Text (OCR) tool handles both approaches — just upload your file and get the text instantly.