Recipe Box Digitizer from Handwritten Cards

WHAT IT DOES

This userscript helps you digitize handwritten recipe cards by extracting structured data from any text source. It scans any webpage containing text from handwritten recipe cards—scanned images with OCR text, recipe blogs, or manual text entry—for recipe structure: title, ingredients, instructions, prep time, cook time, servings, and notes. It parses handwritten-style ingredient lists with quantities and units (e.g., "2 cups flour" → {quantity: 2, unit: "cup", ingredient: "flour"}), splits instructions into numbered steps, and auto-detects measurement systems. The script scales recipes up or down by any factor, preserves the original card's character with optional styling, and exports as plain text, JSON, or Markdown. It also supports batch digitization of multiple cards.

WHERE IT RUNS

The script works on any webpage with recipe text—scanned card images with OCR, recipe blogs, family websites, or pasted text. It runs entirely within your browser, reading only the visible content of the current page. It does not track your browsing history, access other tabs, or send any data externally. All recipe data is processed locally.

HOW TO USE

After installation, open the panel by clicking the tool's icon in the BotGentz sidebar. Paste a recipe into the text input or click "Extract from page" to scan the current webpage. Click "Extract from input" to process the pasted text. The extracted recipe appears in the preview panel showing title, servings, times, ingredients, instructions, and notes. Use the format dropdown to choose your export format—Plain Text, Markdown, or JSON. Toggle "Include notes" to include recipe notes in the export. Click "Export recipe" to download the structured recipe file. Use the scale dropdown to multiply the recipe by 0.5x, 1x, 1.5x, 2x, or 3x, then click "Scale recipe" to update the preview.

THE TECHNICAL MECHANISM

This script uses a multi-stage extraction pipeline to parse recipe text from handwritten cards. First, it splits the text into lines and scans for section headers using regex patterns for "ingredients", "instructions", "prep time", "cook time", "servings", and "notes". It extracts the title from the first line. Ingredients are parsed line by line using regex to extract quantities (including fractions like "1/2"), units (tsp, tbsp, c, oz, lb, g, kg, ml, l, fl oz, pt, qt, gal), and ingredient names. The script normalizes units to full names using a built-in abbreviation dictionary. Instructions are split by numbered steps (e.g., "1. Preheat oven...") or bullet points, and formatted as a numbered list. The scaling function multiplies all ingredient quantities by a factor while preserving units. The export function formats the recipe as plain text, Markdown, or JSON with proper structure.

THE PANEL

The panel is draggable anywhere on the screen and remembers its position between sessions. It can be collapsed to a small floating bubble to minimize distraction, and expanded again with a single click. The panel's width can be adjusted, and pressing the Escape key closes it instantly. All settings—export format, include notes, and scale factor—are persisted locally using GM_setValue, so your preferences remain intact across page reloads and browser restarts.

PLEASE NOTE

This userscript requires the free BotGentz extension to function. The script is designed for recipe text extraction; it works best on well-structured text from OCR or manual entry. It may not work perfectly on poorly formatted text with inconsistent layout. The ingredient parsing handles common formats but may not capture every variation. Always review the extracted recipe for accuracy. The script does not access external APIs or store any data outside your browser. No data from your browsing is stored or transmitted anywhere—all processing occurs entirely in your browser.