full-stack-skills OCRmyPDF — Core OCR Guide
Overview
OCRmyPDF adds an OCR text layer to scanned PDF files, allowing them to be searched or copy-pasted. It uses Tesseract OCR, supports 100+ languages, produces PDF/A by default, and distributes work across all CPU cores.
For image processing (deskew, rotate, clean), see the ocrmypdf-image skill. For optimization and PDF/A options, see ocrmypdf-optimize. For batch/Docker/scripting, see ocrmypdf-batch. For Python API and plugins, see ocrmypdf-api.
Installation
One-liner installs (recommended)
| OS | Command |
|---|---|
| Debian / Ubuntu | apt install ocrmypdf |
| Fedora | dnf install ocrmypdf tesseract-osd |
| macOS (Homebrew) | brew install ocrmypdf |
| macOS (MacPorts) | port install ocrmypdf |
| FreeBSD | pkg install py-ocrmypdf |
| Snap | snap install ocrmypdf |
pip install (latest version)
# After installing system dependencies (Tesseract, Ghostscript)
pip install ocrmypdf
Verify
ocrmypdf --version
ocrmypdf --help
Requirements
- Python 3.11+
- Tesseract 4.1.1+ (OCR engine)
- Ghostscript 9.54+ or pypdfium2 (PDF rasterization)
- Optional: jbig2enc (compression), pngquant (image optimization), unpaper (cleaning)
Quick Start
# Basic OCR — input scanned PDF, output searchable PDF/A
ocrmypdf input.pdf output.pdf
# OCR an image file directly
ocrmypdf --image-dpi 300 scan.png output.pdf
# OCR in place (only overwrites on success)
ocrmypdf myfile.pdf myfile.pdf
Language Support
OCRmyPDF uses Tesseract language packs. Install them for your OS:
# Debian / Ubuntu
apt-cache search tesseract-ocr # List all language packs
apt install tesseract-ocr-chi-sim # Chinese Simplified
apt install tesseract-ocr-fra # French
# macOS (Homebrew)
brew install tesseract-lang # All languages
# Fedora
dnf search tesseract-langpack
dnf install tesseract-langpack-ita # Italian
Using languages
# Single language
ocrmypdf -l fra document.pdf output.pdf
# Multiple languages
ocrmypdf -l eng+fra bilingual.pdf output.pdf
# Chinese Simplified + English
ocrmypdf -l chi_sim+eng chinese-doc.pdf output.pdf
Note: Use ISO 639-3 codes for language identifiers.
OCR Modes
Default mode (skip existing text)
# Skip pages that already have text — only OCR pages without text
ocrmypdf input.pdf output.pdf
Force OCR (--force-ocr or -m force)
# Rasterize and OCR all pages, even those with existing text
ocrmypdf --force-ocr input.pdf output.pdf
# v17+ short form:
ocrmypdf -m force input.pdf output.pdf
Redo OCR (--redo-ocr or -m redo)
# Replace existing OCR without rasterizing (preserves quality)
ocrmypdf --redo-ocr input.pdf output.pdf
# v17+ short form:
ocrmypdf -m redo input.pdf output.pdf
Skip text (--skip-text or -m skip)
# Skip pages with any text, only OCR blank/image pages
ocrmypdf --skip-text input.pdf output.pdf
# v17+ short form:
ocrmypdf -m skip input.pdf output.pdf
No OCR (image processing only)
# Apply image processing / PDF/A conversion without OCR
ocrmypdf --ocr-engine none input.pdf output.pdf
Page Selection
# OCR only specific pages
ocrmypdf --pages 1,3,5-10 input.pdf output.pdf
# OCR only the first page, minimal changes elsewhere
ocrmypdf --pages 1 --output-type pdf --optimize 0 input.pdf output.pdf
Output Types
# PDF/A (default) — for archival
ocrmypdf --output-type pdfa input.pdf output.pdf
# Standard PDF
ocrmypdf --output-type pdf input.pdf output.pdf
# Auto (v17+) — speculative PDF/A, falls back to standard PDF
ocrmypdf --output-type auto input.pdf output.pdf
# No output PDF — only produce sidecar text
ocrmypdf --output-type none --sidecar text.txt input.pdf -
Sidecar Text File
# Produce a companion text file with OCR text
ocrmypdf --sidecar output.txt input.pdf output.pdf
Metadata
# Set output PDF metadata
ocrmypdf --title "My Document" --author "Author Name" --subject "Subject" input.pdf output.pdf
Parallel Processing
# Use 4 CPU cores (default: all available)
ocrmypdf --jobs 4 input.pdf output.pdf
# Single-threaded
ocrmypdf --jobs 1 input.pdf output.pdf
Common Recipes
Make a scanned PDF searchable
ocrmypdf scanned.pdf searchable.pdf
Convert image to searchable PDF
ocrmypdf --image-dpi 300 scan.jpg output.pdf
OCR a multilingual document
ocrmypdf -l eng+deu+fra multilingual.pdf output.pdf
Re-OCR with newer Tesseract
ocrmypdf --redo-ocr old-ocr.pdf updated.pdf
Strip all text/OCR from a PDF
ocrmypdf --ocr-engine none --force-ocr input.pdf stripped.pdf
Quick Reference
| Task | Command |
|---|---|
| Basic OCR | ocrmypdf input.pdf output.pdf |
| Specify language | ocrmypdf -l fra input.pdf output.pdf |
| Multiple languages | ocrmypdf -l eng+fra input.pdf output.pdf |
| Force re-OCR all pages | ocrmypdf --force-ocr input.pdf output.pdf |
| Replace existing OCR | ocrmypdf --redo-ocr input.pdf output.pdf |
| Skip pages with text | ocrmypdf --skip-text input.pdf output.pdf |
| Specific pages only | ocrmypdf --pages 1,3,5-10 input.pdf output.pdf |
| Output standard PDF | ocrmypdf --output-type pdf input.pdf output.pdf |
| Extract text sidecar | ocrmypdf --sidecar text.txt input.pdf output.pdf |
| Image to PDF | ocrmypdf --image-dpi 300 image.png output.pdf |
| In-place OCR | ocrmypdf myfile.pdf myfile.pdf |
| Set metadata | ocrmypdf --title "Title" input.pdf output.pdf |
| Parallel jobs | ocrmypdf --jobs 4 input.pdf output.pdf |
Troubleshooting
- "Tesseract not found": Install Tesseract and ensure it's on PATH.
- Poor OCR quality: Check language packs (
-l), try--deskew(see ocrmypdf-image), or--oversample 300. - "Input file has text": Use
--force-ocr,--redo-ocr, or--skip-textas appropriate. - Large output files: See ocrmypdf-optimize for
--optimizelevels and JBIG2. - Signed PDFs: Use
--invalidate-digital-signaturesto override (signatures will be invalidated).
References
国内适配
- 支持中文文档和中文注释
- 示例代码兼容国内开发环境
- 提供中文 FAQ 和常见问题解答
能力边界
✅ 适用场景
- 当你需要使用此技能对应的技术栈时
- 当项目需要遵循最佳实践时
- 当需要快速上手或深入理解核心概念时
⚠️ 需要注意
- 复杂业务逻辑需要结合具体场景调整
- 性能优化需要根据实际数据量评估
❌ 不适用场景
- 不相关的技术栈或框架
- 需要完全自定义的特殊场景
使用流程
Step 1: 环境准备
确保开发环境已安装必要的依赖和工具。
Step 2: 配置初始化
根据项目需求进行基础配置。
Step 3: 核心功能使用
按照示例代码实现核心功能。
Step 4: 测试验证
运行测试确保功能正常。
Step 5: 部署上线
完成开发后进行部署和监控。