Handwriting and Image to LaTeX

Handwriting and Image to LaTeX

Handwriting and Image to LaTeX is an Obsidian plugin that lets you draw mathematical formulas directly on a canvas, or import images from your gallery, and convert them into LaTeX code using optical character recognition (OCR). It supports multiple API backends, with SimpleTex as the recommended (and free) option.

---

✨ Features

• ✍️ Handwriting canvas – draw with your finger or mouse, with pen/eraser/pan modes.
• 📷 Image import – load images from your device’s photo gallery.
• 🧠 Formula recognition – powered by:
  • SimpleTex (recommended, free UAT token)
  • Custom APIs (JSON or multipart/form-data)
  • (Experimental) Mathpix, OpenAI GPT-4o Vision
• 🎯 Crop window – select exactly the region you want to recognize.
• 🌀 Infinite canvas (rewritten) – viewport-based infinite canvas in the style of Tldraw/Excalidraw; strokes live in world coordinates and only visible content is rendered, so large drawings stay fast.
• 🔲 Selection recognition – rectangle, lasso, or tap-to-add polygon selection. Strokes are treated as whole objects: any stroke intersecting a selection is included in full, and each selected group is recognized independently into its own result slot. Selected strokes get a subtle accent halo. Recognition history stores canvas snapshots, so back/forward restores both result and drawing.
• ✂️ Smart crop – automatic tight cropping before upload, plus image rotation and enhancement filters (auto enhance, grayscale, sharpen, binarize) in the crop window.
• 📐 Math mode – insert as inline ($...$), display ($$...$$), or raw LaTeX.
• 📱 Mobile friendly – fully optimized for touch devices.
• 🎨 Customizable – stroke color/width, grid, default canvas size.

---

Sample images

---

📦 Installation

From Obsidian Community Plugins

1. Open Settings → Community Plugins.
2. Disable Safe Mode.
3. Click Browse and search for “Handwriting to LaTeX”.
4. Install and enable the plugin.

Manual (BRAT)

1. Install the BRAT plugin.
2. Add this repository: https://github.com/ChongCyrus/handwriting-latex.
3. Enable the plugin.

---

🚀 Usage

Open the handwriting modal

• Click the pencil icon in the ribbon (left sidebar).
• Or run the command: “Open handwriting canvas” from the Command Palette (Cmd/Ctrl+P).

Drawing on the canvas

• Pen mode (✏️) – draw freely.
• Eraser mode (🧹) – tap/click strokes to remove them.
• Hand mode (✋) – pan around the canvas.
• Rect select (▭) – drag a rectangle to select strokes for recognition.
• Lasso select (◯) – draw a freehand circle to select strokes.
• Polygon select (🔷) – tap to place vertices; tap the first vertex again to close the polygon.
• Clear selections (❎) – clear all selection regions.
• Zoom – pinch on touch, or use mouse wheel.
• Undo (↩️) – remove the last stroke.
• Clear (🗑️) – erase everything.
• Rotate (↺/↻) – rotate all strokes by 90°.
• Fit view (⊘) – zoom to fit all content.
• Reset view (🏠) – return to default zoom/pan.
• Grid toggle (⊞) – show/hide background grid.

Recognizing a formula

1. Draw your formula (or import an image via the 📷 or 🖼️ buttons).
2. Optionally use ▭ / ◯ / 🔷 to select one or more regions; each selection is recognized independently.
3. Click the “🔍 Recognize” button (top‑right or inside the crop window).
4. Each recognition writes into its own result card; use the card’s “✅ Insert” button, or select a result in the preview and use the header Insert button.
5. Choose your insert mode (inline/display/raw).

Importing images

• Camera (📷) – experimental; opens the device camera when Obsidian has camera permissions. Many mobile builds block this, so prefer Album.
• Album (🖼️) – opens your file picker to select an image.
• After importing, you can rotate, enhance, and crop the image to the formula area before recognition.

---

⚙️ Configuration

Go to Settings → Handwriting to LaTeX to configure:

---

🔐 API Setup Guides

SimpleTex (Recommended)

1. Register at SimpleTex.
2. Go to User Center → User Authorization Token.
3. Create a token (UAT) and copy it.
4. Paste it into the API Key field.

Custom API (JSON)

• Your endpoint should accept a POST request with JSON body:

  {
    "image": "data:image/png;base64,...",
    "prompt": "your prompt"
  }