A localized Japanese OCR solution for high-accuracy text extraction from vertical and horizontal printed documents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install jpocr
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 jpocr 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 jpocr?
jpocr is a specialized Japanese Optical Character Recognition (OCR) skill built for the Openclaw Skills ecosystem. It utilizes NDLOCR-Lite, a lightweight version of the National Diet Library's OCR engine, to provide industry-leading accuracy for printed Japanese text without the need for a GPU or external API keys. This skill is a vital component for developers needing to process sensitive documents locally while maintaining high performance on modern CPU architectures like Apple Silicon.
By prioritizing local execution, jpocr ensures data privacy and eliminates latency issues associated with cloud-based OCR services. It is particularly effective at handling complex layouts, including traditional vertical Japanese text (tategaki), making it a versatile choice for archival and digitization projects within the Openclaw Skills environment.
jpocr Use Cases
- Automating text extraction from scanned Japanese historical documents.
- Converting vertical text screenshots into editable digital formats.
- Batch processing image folders for searchable database integration using Openclaw Skills.
- Rapid digitization of printed Japanese manuals and books without cloud dependency.
How jpocr Works
- The user invokes the skill via the CLI, passing an image path or directory to the script.
- The system employs the DEIMv2 layout detection model to map out text regions within the document.
- The PARSeq cascade recognition engine analyzes the text regions using tiered 30/50/100 character models.
- An xy-cut algorithm determines the logical reading order, correctly identifying horizontal and vertical flows.
- The final output is generated in the requested format, such as plain text or structured JSON data.
jpocr Setup
To integrate this skill into your local Openclaw Skills workflow, navigate to the skill directory and use the provided shell scripts. Ensure you have a functional bash environment on macOS or Linux.
# Basic text extraction from an image
<SKILL_ROOT>/scripts/ocr-cli.sh path/to/image.jpg
# Advanced usage with JSON output for data processing
<SKILL_ROOT>/scripts/ocr-cli.sh path/to/image.jpg --json
jpocr Data Schema & Taxonomy
The skill organizes its output into structured formats designed for easy ingestion by other Openclaw Skills.
| Property | Type | Description |
|---|---|---|
| text | String | The recognized characters from the detected region. |
| boundingBox | Array | Four [x, y] coordinates defining the text area. |
| confidence | Float | The model's certainty score for the recognized text. |
| isVertical | Boolean | Indicates if the text region is vertically oriented. |
The JSON output also includes an imginfo object containing the original dimensions (width and height) of the processed file for spatial context.
jpocr Advanced Features
- Optimized for Apple Silicon (M-series) and x86 CPU performance using ONNX runtimes.
- Support for a wide range of image formats including JP2, TIFF, BMP, JPG, and PNG.
- Advanced visualization mode to verify layout detection accuracy by saving bounding-box overlays.
- Extensible character set covering approximately 7,000 characters including JIS kanji, kana, and ASCII.
- Fully offline operation ensures the absolute security of documents processed via Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
TikTok Viral Marketing
A specialized AI agent skill for sourcing and managing TikTok creators to execute viral marketing campaigns and trending challenges.
TikTok UGC Creator
A programmatic interface for AI agents to hire authentic TikTok creators for high-trust user-generated content and testimonials.
TikTok Trend Challenger
An AI agent skill designed to identify and hire TikTok creators for participation in viral trends and hashtag challenges to maximize algorithmic reach.
TikTok Product Promotion
A powerful AI agent skill for sourcing and managing TikTok influencers to create high-impact product reviews, unboxing videos, and conversion-driven campaigns.
Zoomin Scraper
An automated browser-based scraper specifically designed to extract dynamic documentation content from Zoomin Software portals.
Earl Display Control
A comprehensive control suite for managing the VisuoSpatial Sketchpad dashboard, including server management, kiosk browser deployment, and agent state updates.
Featured*
