OpenMathpix

Open-source desktop snipping tool that captures screen regions and converts math formulas, tables, text into LaTeX, Markdown, or plain text — powered by PaddleOCR on Baidu AIStudio.

---

Installation

Windows (portable)

1. Download OpenMathpix-win-portable.zip from Releases.
2. Extract the zip file.
3. Run OpenMathpix.exe from the extracted folder.

Build from source

git clone git@github.com:JhuoW/OpenMathpix.git
cd OpenMathpix
npm install
npm run dev        
npm run package    

Getting started

Step 1: Get your API credentials

1. Go to https://aistudio.baidu.com/paddleocr and log in (or create a free Baidu AIStudio account).
2. Click "API" button to access the model list. Then select a model (e.g. PaddleOCR-VL) to reveal the example code and your unique API URL and Access Token.

3. Copy these two values from the example:

• API URL — your unique endpoint, e.g. https://xxxxxx.aistudio-app.com/layout-parsing
• Access Token — your authentication token (also at https://aistudio.baidu.com/index/accessToken)

Note: Each model gets its own API URL (different subdomain). If you switch models, copy the new URL too.

Step 2: Configure OpenMathpix

1. Launch OpenMathpix.
2. Click Settings in the toolbar.
3. Under API Configuration:
   • Paste your API URL.
   • Paste your Access Token.
   • Select a Pipeline that matches the model you chose on AIStudio.
4. Click Test Connection to verify.

Step 3: Capture and recognize

Screen snip

You can click Snip button in the toolbar and then click New Snip button to dim the screen.

You can click and drag to select a region. Release to capture.

Viewing results

After recognition, results appear in four tabs:

Tab	Description
LaTeX	Extracted math rendered via KaTeX, with copyable raw source.
Markdown	Structured output (headings, tables, inline math).
Text	Plain text with all formatting stripped.
Image	The original captured screenshot.

Click Copy to copy the active tab. The default format is auto-copied after each recognition.

Pipelines

Pipeline	Best for	Output
PP-OCRv5	General text (CJK + English)	Text only
PP-StructureV3	Math, tables, charts, mixed layout	Markdown + LaTeX
PaddleOCR-VL	Vision-language document parsing	Markdown + LaTeX
PaddleOCR-VL-1.5	Higher accuracy, seal & irregular text	Markdown + LaTeX

Each pipeline has its own API URL on AIStudio. When switching pipelines, update the URL in Settings to match.

Settings

Setting	Default	Description
API URL	—	Your PaddleOCR AIStudio endpoint.
Access Token	—	Stored encrypted on disk via safeStorage.
Pipeline	PP-StructureV3	Which OCR pipeline to use.
Snip Hotkey	Ctrl+Shift+S	Global shortcut for screen capture.
Default Output	LaTeX	Format auto-copied after recognition.
Theme	System	Light / Dark / System.
History Limit	100	Max saved recognitions (10–500).

Self-hosted: Set the API URL to your local server (e.g. http://localhost:8080) and leave the token empty.

History

Click History in the toolbar to browse past recognitions. Each entry shows a thumbnail, timestamp, and preview. Use the search bar to filter, or click an entry to re-open its result.

System tray

Closing the window minimizes to tray. Right-click the tray icon for:

• Snip — start capture
• Open Window — show main window
• Settings — open settings
• Quit — exit

Keyboard shortcuts

Shortcut	Action
Ctrl+Shift+S	Screen snip (global)
Ctrl+V	Paste image for recognition
Escape	Cancel snip

License

MIT