Skip to content

Marker Converters

Complete reference for all converter types in marker-pdf.

Converter Types

PdfConverter (Default)

Full-featured PDF converter with OCR, layout detection, and table recognition.

from marker.converters.pdf import PdfConverter
from marker.models import create_model_dict

converter = PdfConverter(artifact_dict=create_model_dict())
rendered = converter("document.pdf")

Features: - Text extraction from native PDFs - OCR for scanned pages - Table detection and extraction - Equation recognition - Image extraction - Layout preservation

Best for: General PDF conversion, mixed content documents

TableConverter

Specialized converter for extracting tables from PDFs.

from marker.converters.table import TableConverter
from marker.models import create_model_dict

converter = TableConverter(artifact_dict=create_model_dict())
rendered = converter("document.pdf")

Features: - Focused table detection - Structured table output - Cell-level extraction - Merged cell handling

Best for: Documents with many tables, data extraction

Output format: JSON with table structures

{
  "tables": [
    {
      "page": 1,
      "bbox": [x1, y1, x2, y2],
      "cells": [
        {"row": 0, "col": 0, "text": "Header 1"},
        {"row": 0, "col": 1, "text": "Header 2"}
      ]
    }
  ]
}

OCRConverter

Pure OCR converter without layout analysis.

from marker.converters.ocr import OCRConverter
from marker.models import create_model_dict

converter = OCRConverter(artifact_dict=create_model_dict())
rendered = converter("document.pdf")

Features: - Raw OCR text extraction - No layout preservation - Fast processing - Simple output

Best for: Scanned documents, text-only extraction, speed priority

Output: Plain text in reading order

EquationConverter

Specialized converter for mathematical equations.

from marker.converters.equation import EquationConverter
from marker.models import create_model_dict

converter = EquationConverter(artifact_dict=create_model_dict())
rendered = converter("document.pdf")

Features: - LaTeX equation extraction - Inline and display equations - Symbol recognition - Math notation preservation

Best for: Academic papers, mathematical documents

Output format: LaTeX equations

$$
E = mc^2
$$

Inline equation: $x^2 + y^2 = z^2$

Converter Comparison

Converter Speed Accuracy Layout Tables Equations Images
PdfConverter Medium High ✓ ✓ ✓ ✓
TableConverter Fast High ✗ ✓ ✗ ✗
OCRConverter Fast Medium ✗ ✗ ✗ ✗
EquationConverter Medium High ✗ ✗ ✓ ✗

Using Converters via CLI

PdfConverter (default)

marker_single document.pdf

TableConverter

marker_single document.pdf \
  --converter_cls marker.converters.table.TableConverter \
  --output_format json

OCRConverter

marker_single document.pdf \
  --converter_cls marker.converters.ocr.OCRConverter

EquationConverter

marker_single document.pdf \
  --converter_cls marker.converters.equation.EquationConverter

Custom Converters

Create custom converters by extending BaseConverter:

from marker.converters.base import BaseConverter
from marker.models import create_model_dict

class CustomConverter(BaseConverter):
    def __init__(self, artifact_dict, **kwargs):
        super().__init__(artifact_dict, **kwargs)
        # Custom initialization

    def convert_page(self, page):
        # Custom page conversion logic
        return processed_page

    def post_process(self, pages):
        # Custom post-processing
        return final_output

# Usage
converter = CustomConverter(artifact_dict=create_model_dict())
rendered = converter("document.pdf")

Converter Configuration

Common Options

All converters support these configuration options:

config = {
    "page_range": (0, 10),      # Process pages 0-10
    "languages": ["en", "es"],  # Expected languages
    "workers": 4,               # Parallel workers
    "output_format": "markdown" # Output format
}

converter = PdfConverter(
    artifact_dict=create_model_dict(),
    config=config
)

Converter-Specific Options

PdfConverter

config = {
    "use_llm": True,           # Use LLM for accuracy
    "force_ocr": False,        # Force OCR on all pages
    "extract_images": True,    # Extract embedded images
    "preserve_layout": True    # Maintain document layout
}

TableConverter

config = {
    "min_table_confidence": 0.8,  # Minimum confidence threshold
    "merge_cells": True,           # Handle merged cells
    "extract_headers": True        # Identify table headers
}

OCRConverter

config = {
    "ocr_confidence": 0.7,     # Minimum OCR confidence
    "denoise": True,           # Apply denoising
    "deskew": True            # Correct page skew
}

Chaining Converters

Process documents with multiple converters:

from marker.converters.pdf import PdfConverter
from marker.converters.table import TableConverter
from marker.models import create_model_dict

models = create_model_dict()

# Extract general content
pdf_converter = PdfConverter(artifact_dict=models)
text, metadata, images = text_from_rendered(pdf_converter("doc.pdf"))

# Extract tables separately
table_converter = TableConverter(artifact_dict=models)
tables = text_from_rendered(table_converter("doc.pdf"), output_format="json")

# Combine results
combined = {
    "text": text,
    "tables": tables,
    "metadata": metadata
}

See Also