Getting the Best Results from AI Import
Garde’s AI recipe import handles a variety of input formats, but the quality of your source material directly affects the accuracy of the results. These tips will help you get clean, accurate imports with fewer corrections.Photo Quality Tips
If you are uploading photos of recipe cards, binders, or prep sheets, keep the following in mind:Lighting
- Use natural or bright overhead lighting. Avoid harsh shadows that fall across the text.
- Do not use flash if it creates glare on laminated or glossy surfaces.
Angle and Focus
- Hold your camera directly above the document, parallel to the surface, to avoid distortion.
- Make sure the text is in sharp focus before taking the photo. Blurry images are the number one cause of extraction errors.
- Capture the entire recipe in one photo when possible. If the recipe spans multiple pages, take one photo per page.
Background
- Place the recipe on a clean, contrasting background. A white recipe card on a white counter is harder for AI to parse than a white card on a dark surface.
- Remove any clutter or overlapping documents from the frame.
Handwritten Recipes
- AI works best with printed or typed text. Handwritten recipes can be processed, but accuracy depends on legibility.
- If your handwritten recipes are difficult to read, consider typing them into a simple spreadsheet before uploading. This often saves more time than correcting extraction errors.
PDF and Document Tips
- Use text-based PDFs when possible. PDFs where you can select and copy text are processed much more accurately than scanned images.
- Check for clean formatting. Recipes with clear headers, ingredient lists, and separated sections give AI the best structure to work with.
- Remove extraneous pages. If your PDF contains a mix of recipes and other content (like order guides or equipment lists), trim it to just the recipe pages before uploading.
Spreadsheet Tips
- Use one row per ingredient. Put the ingredient name, quantity, and unit in separate columns.
- Include a recipe name column or header row so AI can identify which recipe each ingredient belongs to.
- Keep units consistent. If you use “oz” in some places and “ounces” in others, AI can usually handle this, but consistency reduces the chance of errors.
- Separate multiple recipes clearly. If your spreadsheet contains multiple recipes, use clear breaks or separate tabs for each recipe.
Handling Complex Recipes
Recipes with Sub-Recipes
Many restaurant recipes include components that are themselves recipes (for example, a burger recipe that includes a house-made aioli). When importing these:- Upload the sub-recipe (aioli) first and confirm it in Garde.
- Then upload the parent recipe (burger). During review, you can link the sub-recipe as an ingredient instead of listing its individual components.
- If you upload everything at once, Garde will flag potential sub-recipe relationships for your review.
Recipes with Variable Yields
Some recipes have different yield amounts depending on batch size. When uploading:- Choose one standard batch size to import.
- You can adjust the yield after import. Garde will scale ingredient quantities proportionally.
Recipes with Prep Instructions Only
If your recipe document contains only preparation steps without a clear ingredient list, AI may not be able to extract structured ingredient data. In these cases:- Add a simple ingredient list to the document before uploading.
- Or, create the recipe manually in Garde and use the instructions field for the prep steps.
Reviewing and Correcting AI Suggestions
After AI processes your upload, take time to review each recipe carefully:Check Quantities and Units
This is where errors are most likely. Common issues include:- Misread numbers — A handwritten “7” might be read as “1” or vice versa.
- Unit confusion — “lb” might be read as “lb” (correct) or “16” (incorrect). Double-check that units make sense for each ingredient.
- Decimal placement — Verify that “1.5 oz” was not read as “15 oz.”
Verify Product Matches
When AI suggests a product match from your database, confirm that it selected the right product. Watch for:- Generic vs. specific matches — AI might match “tomatoes” to canned tomatoes when you meant fresh roma tomatoes.
- Brand-specific products — If you use a specific brand, make sure the match reflects that.