Skip to main content
Mizen can parse recipes directly from images using AI vision technology. Upload photos of recipe cards, cookbook pages, or screenshots to extract structured recipe data.

​
Supported Image Types

Mizen accepts common image formats:
  • JPG/JPEG - Standard photo format
  • PNG - Screenshots and graphics
  • WEBP - Modern web format
  • GIF - Animated or static images
Maximum file size: 10MB Format validation in /home/daytona/workspace/source/src/app/api/parseRecipeFromImage/route.ts:61-79.

​
How to Use

1

Switch to image mode

Click the dropdown arrow in the search bar and select the image upload option.
2

Select an image

Click the upload area or drag and drop an image file. A preview will appear.
3

Parse the recipe

Click “Parse Recipe from Image” to begin processing. This typically takes 3-5 seconds.
4

Review results

The parsed recipe appears in the recipe viewer with all extracted data.

​
What Gets Extracted

The AI vision model extracts:
  • Title - Recipe name from image
  • Ingredients - Complete ingredient list with:
    • Amounts (quantities)
    • Units (cups, tablespoons, etc.)
    • Ingredient names
    • Grouped sections (if present)
  • Instructions - Step-by-step directions
  • Metadata - When visible:
    • Servings/yield
    • Prep time
    • Cook time
    • Author
    • Cuisine type
  • Summary - AI-generated recipe overview
Implemented in /home/daytona/workspace/source/src/app/api/parseRecipeFromImage/route.ts:44-226.

​
Processing Pipeline

​
Client-Side

1

File validation

Check file type and size before uploading.
2

Preview generation

Convert image to base64 data URL for preview.
3

Upload to API

Send image as multipart/form-data.
Client logic in /home/daytona/workspace/source/src/components/search/recipe-search-form.tsx:122-155.

​
Server-Side

1

Receive upload

Parse multipart/form-data and extract image file.
2

Convert to base64

Transform image to base64 data URL for AI model.
3

Vision API call

Send to AI vision model (Groq) for analysis.
4

Extract structured data

Parse AI response into standard recipe format.
5

Return results

Send structured recipe data back to client.
Server processing in /home/daytona/workspace/source/src/app/api/parseRecipeFromImage/route.ts:102-119.

​
Image Requirements

For best results, ensure images:

​
Are Clear and Readable

  • Text should be in focus
  • Adequate lighting and contrast
  • Minimal glare or shadows
  • Straight orientation (not rotated)

​
Contain Complete Information

  • Full ingredient list visible
  • All instruction steps shown
  • No text cut off at edges

​
Have Good Resolution

  • Minimum 800x600 recommended
  • Higher resolution improves accuracy
  • Avoid heavily compressed images
For multi-page recipes, combine pages into a single image or parse each page separately, then manually combine the results.

​
Error Handling

Common errors and solutions:

​
ERR_INVALID_FILE_TYPE

Cause: Unsupported file format Solution: Convert to JPG, PNG, WEBP, or GIF

​
ERR_FILE_TOO_LARGE

Cause: Image exceeds 10MB Solution: Compress or resize the image

​
ERR_NO_RECIPE_FOUND

Cause: AI couldn’t extract recipe text Solutions:
  • Improve image quality
  • Ensure recipe text is visible
  • Try a different image

​
ERR_RATE_LIMIT

Cause: Too many requests Solution: Wait before trying again (retry-after timestamp provided) Error codes in /home/daytona/workspace/source/src/app/api/parseRecipeFromImage/route.ts:129-161.

​
Response Format

{
  "success": true,
  "title": "Chocolate Chip Cookies",
  "ingredients": [
    {
      "groupName": "Main",
      "ingredients": [
        {
          "amount": "2",
          "units": "cups",
          "ingredient": "flour"
        }
      ]
    }
  ],
  "instructions": [
    "Preheat oven to 350°F",
    "Mix dry ingredients"
  ],
  "imageData": "data:image/jpeg;base64,...",
  "imageFilename": "recipe-card.jpg"
}

​
Image Storage

Parsed images are stored:
  • Base64 in localStorage - For recipe preview
  • Filename preserved - Original name retained
  • Source reference - Stored as image:filename
This enables:
  • Recipe thumbnail display
  • Source tracking
  • Re-access without re-upload
Images are stored locally in your browser. They are not uploaded to external servers beyond the initial parsing request.

​
AI Model

Mizen uses Groq’s vision API for image parsing:
  • Provider: Groq
  • Model: Vision-enabled model
  • Input: Base64-encoded images
  • Output: Structured JSON recipe data

​
Limitations

​
Text Recognition

  • Handwritten recipes may have lower accuracy
  • Unusual fonts may cause parsing issues
  • Decorative text can confuse the AI

​
Image Quality

  • Blurry images reduce accuracy
  • Low contrast makes text harder to read
  • Heavy compression loses detail

​
Language Support

  • Optimized for English recipes
  • Other languages may work but with reduced accuracy

​
API Dependencies

  • Requires valid Groq API key
  • Subject to API rate limits
  • Network connection required
For frequently used recipes, consider re-parsing from the original URL instead of images when possible. URL parsing is faster and more accurate.

​
Performance

  • Typical parse time: 3-5 seconds
  • Network upload: ~1 second (depends on image size)
  • AI processing: ~2-4 seconds
  • Timeout: 2 minutes maximum

​
Privacy

Image parsing privacy considerations:
  • Images sent to Groq API for processing
  • No permanent storage on Mizen servers
  • Base64 data stored in browser localStorage only
  • Can be cleared by user at any time

Build docs developers (and LLMs) love

Get started for freeTalk to us