An intelligent OCR and document parsing engine that extracts text, mathematical formulas, and tables from images and PDFs with high precision.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install opencr-skill
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install opencr-skill using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is OpenOCR Skill?
OpenOCR is a comprehensive and efficient general OCR system designed to handle a wide variety of visual recognition tasks. As a key component in the Openclaw Skills ecosystem, it provides developers with a unified interface for text detection, recognition, and sophisticated document parsing. By leveraging Vision Language Models (VLM), it excels at recognizing complex structures like mathematical formulas and tables that traditional OCR systems often struggle with.
This skill is built for versatility, offering both lightweight mobile modes for speed and server-grade modes for maximum accuracy. Whether you are processing a single screenshot or a multi-page scanned PDF, OpenOCR provides the tools necessary to transform visual data into structured, machine-readable formats like Markdown or JSON, enhancing the capabilities of any Openclaw Skills implementation.
OpenOCR Skill Use Cases
- Converting scanned PDF reports into structured Markdown for easier indexing and search.
- Extracting LaTeX-formatted mathematical formulas from technical papers or screenshots.
- Automating the digitization of printed tables into JSON data for analysis.
- Performing layout analysis on complex documents to identify headings, paragraphs, and image captions.
- Building automated data entry pipelines that process images from a directory using Openclaw Skills.
How OpenOCR Skill Works
- Input ingestion: The skill accepts image files (JPG, PNG, BMP) or PDF documents as the primary source.
- Task Selection: Users define the specific operation, such as detection (det), recognition (rec), end-to-end OCR (ocr), or universal recognition (unirec).
- Layout Analysis: For document tasks, the engine uses PP-DocLayoutV2 to identify different content blocks like text, tables, and charts.
- Recognition Pipeline: The chosen backend (ONNX or PyTorch) processes the visual data to extract text strings, formula LaTeX, or table structures.
- Result Serialization: The processed data is returned as a structured object, which can be saved as Markdown, JSON, or visualized with bounding boxes.
OpenOCR Skill Setup
To begin using this skill within your Openclaw Skills environment, install the package via pip. You can choose the basic installation or include optional dependencies for GPU support.
# Basic installation (CPU, ONNX backend)
pip install openocr-python
# GPU-accelerated ONNX inference
pip install openocr-python[onnx-gpu]
# PyTorch backend for high-accuracy server mode
pip install openocr-python[pytorch]
# Install all optional dependencies including Gradio demos
pip install openocr-python[all]
OpenOCR Skill Data Schema & Taxonomy
OpenOCR provides detailed metadata for every extraction task. The following table describes how the data is organized when utilizing Openclaw Skills:
| Feature | Data Type | Description |
|---|---|---|
| Boxes | np.ndarray | Bounding box coordinates for detected text regions |
| Text | String | The recognized characters or LaTeX formulas |
| Score | Float | Confidence score (0.0 to 1.0) for the recognition |
| Elapse | Float | Time taken in seconds to process the task |
| Layout Blocks | Dictionary | Structured content identified during document parsing |
OpenOCR Skill Advanced Features
- Universal Recognition (UniRec): Uses a 0.1B parameter VLM to handle text, formulas, and tables in a single pass.
- Multi-Backend Support: Switch seamlessly between ONNX for CPU efficiency and PyTorch for GPU-accelerated accuracy.
- PDF Native Processing: Automatically handles multi-page PDF documents, converting pages to images internally for analysis.
- Layout Analysis: Advanced document parsing that distinguishes between different types of content blocks for better structure retention.
- Visualization Tools: Generate images with overlaid bounding boxes and labels to verify extraction accuracy.
- CLI Integration: Execute complex OCR tasks directly from the command line, facilitating automation within Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
Habitica Gamified Task Management Skill
A comprehensive gamification integration that allows AI agents to manage habits, dailies, and character progression using the Habitica API.
API Gateway
A robust passthrough proxy providing AI agents with direct access to over 100 native APIs via managed OAuth connections.
update-md
A standardized framework for creating and maintaining modular, deployment-synced project documentation using Markdown.
Tonic System Deploy
A sophisticated deployment workflow manager for handling structured UAT and production environments with automated nightly pipelines and human approval gates.
OpenClawdy
A persistent memory infrastructure for autonomous agents that enables semantic recall, reputation tracking, and cross-agent knowledge sharing.
OpenClawdy
OpenClawdy is a memory infrastructure for autonomous agents providing persistent storage, semantic recall, and cross-agent knowledge sharing.
Featured*
