Custom Conversion

Customize PDF conversion with OCR engines, backends, and pipeline options

Docling provides extensive customization options for PDF conversion, allowing you to toggle OCR engines, backends, and pipeline settings.

Overview

This example demonstrates:

• How to configure OCR options (EasyOCR, Tesseract, macOS OCR)
• Switching between PDF backends
• Customizing table structure recognition
• Setting accelerator options for GPU/CPU

Basic Configuration

custom_convert.py

from pathlib import Path
from docling.datamodel.accelerator_options import AcceleratorDevice, AcceleratorOptions
from docling.datamodel.base_models import InputFormat
from docling.datamodel.pipeline_options import (
    PdfPipelineOptions,
    TableStructureOptions,
)
from docling.document_converter import DocumentConverter, PdfFormatOption

input_doc_path = Path("path/to/document.pdf")

# Configure pipeline options
pipeline_options = PdfPipelineOptions()
pipeline_options.do_ocr = True
pipeline_options.do_table_structure = True
pipeline_options.table_structure_options = TableStructureOptions(
    do_cell_matching=True
)
pipeline_options.ocr_options.lang = ["es"]  # Set language
pipeline_options.accelerator_options = AcceleratorOptions(
    num_threads=4, device=AcceleratorDevice.AUTO
)

doc_converter = DocumentConverter(
    format_options={
        InputFormat.PDF: PdfFormatOption(pipeline_options=pipeline_options)
    }
)

result = doc_converter.convert(input_doc_path)

OCR Engine Options

EasyOCR (Default)

pipeline_options = PdfPipelineOptions()
pipeline_options.do_ocr = True
pipeline_options.ocr_options.lang = ["en", "de"]
pipeline_options.do_table_structure = True
pipeline_options.table_structure_options = TableStructureOptions(
    do_cell_matching=True
)

Export Results

1. Convert Document

   Process the document with your custom configuration.

2. Export to Multiple Formats

   Save results as JSON, Markdown, plain text, and doctags.

from pathlib import Path
import json

output_dir = Path("scratch")
output_dir.mkdir(parents=True, exist_ok=True)
doc_filename = result.input.file.stem

# Export to JSON
with (output_dir / f"{doc_filename}.json").open("w") as fp:
    fp.write(json.dumps(result.document.export_to_dict()))

# Export to Markdown
with (output_dir / f"{doc_filename}.md").open("w") as fp:
    fp.write(result.document.export_to_markdown())

# Export to plain text
with (output_dir / f"{doc_filename}.txt").open("w") as fp:
    fp.write(result.document.export_to_markdown(strict_text=True))

# Export to doctags
with (output_dir / f"{doc_filename}.doctags").open("w") as fp:
    fp.write(result.document.export_to_doctags())

Adjust pipeline_options.ocr_options.lang to match your document’s language. Examples: ["en"], ["es"], ["en", "de"].

Accelerator Configuration

Tune performance with accelerator options:

from docling.datamodel.accelerator_options import AcceleratorDevice, AcceleratorOptions

pipeline_options.accelerator_options = AcceleratorOptions(
    num_threads=4,
    device=AcceleratorDevice.AUTO  # or CPU, CUDA, MPS
)