AI Restaurant Menu Parser (No AI Markup) + Allergens

$0.1/1K dishes | Turn menu photos into dish JSON: translated names, prices, ingredients, dietary type and AI-estimated EU-14 allergens and nutrition in 15 languages. Runs on your own Gemini key. Not affiliated with Google.

AI Menu Photo to Dish JSON Parser

Our parser reads photos of a restaurant menu and returns one JSON record per dish: the name in your language, the price as printed, atomic ingredients with EU-14 allergen flags, a dietary type and an estimated nutrition block. It runs on your own Google AI Studio (Gemini) key, so you control the AI spend and the quota. It is ideal for dietary-restriction apps, delivery and menu databases, and menu translation.

• Don't want to manage a Gemini key? Use AI Restaurant Menu Parser (No API Key) + Allergens: same parser, the AI cost is included, $0.20 per 1,000 dishes.
• Mostly dense menus? AI Menu Parser (Flat Price per Menu) + Allergens charges one flat $0.02 per menu of up to 10 photos, however many dishes it holds; it beats the No API Key per-dish price on menus above roughly 100 dishes.
• Have the menu and want dish photos? Dish Photo Matcher (No AI Markup): Food Photos to Menu Items matches guest photos to your menu items.
• Only have a Google Maps place? Our Google Maps Menu Scraper finds the menu photos for you.
• Browse all our scrapers: https://apify.com/traveleat

Key Benefits

• Your key, your cost: parsing runs on your Gemini key, so the AI bill and quota stay under your control (Google bills it separately).
• Low pricing: $0.10 per 1,000 dishes, plus a $0.00005 start fee per run.
• Free to try: Start on the prefilled form returns a clearly marked DEMO row at no charge, and your first live run waives up to 10 dish charges.
• No charge for empty runs: unreadable photos produce a diagnostic row, not a bill.
• Rich output: ingredients, EU-14 allergens, dietary type and estimated nutrition in 15 output languages.

How It Works

1. Add your menu photos: upload JPEG, PNG or WebP files in the Console, or pass image URLs (API callers can also send base64). Up to 20 photos per run are parsed together as one menu, and each image can be up to 10 MiB.
2. Paste your Gemini key (get a free one at aistudio.google.com/apikey) and pick the output language and Gemini model.
3. Run the Actor. Each readable dish becomes one dataset row; an empty or failed parse returns one diagnostic row instead.

More settings:

• geminiModel - the model that reads the photos. Default is gemini-2.5-flash.
• language - output language for names, descriptions and allergen summaries. Default is en; ingredient database names stay English.

The prefilled form (the sample menu photo with no Gemini key) is a deterministic preview: it returns one DEMO row, calls no model and bills nothing. Add your own photos and your key for a live parse.

Limits: PDFs are not read, so convert menu pages to images first. Unreadable items are skipped rather than guessed, and AI output can be wrong on stylised or low-resolution menus.

Allergens, dietary type, ingredients and nutrition are AI estimates. They can be wrong and do not reliably account for substitutions or kitchen cross-contact. Verify critical information with the restaurant, and never use this output as the only safeguard for a severe allergy.

Data You'll Receive

• Dish basics: name translated to your language, originalName as printed, category, price with currency as shown, composition, description.
• Ingredients: atomic base ingredients, each with an English database name, a translated display name and EU-14 allergens with a confidence level (AI estimates).
• Diet and nutrition: dietaryType (Vegan, Vegetarian, Pescetarian or Other) with an explanation, estimated serving size, calories, protein, fat and carbs, isAlcoholic for drinks.
• Story: an origin or cultural note on some dishes (11 of 12 in our test menu).
• Source: sourceImageUrls and sourceImageRefs point back to the photo each dish came from.

Sample output data

Here is an example dish from a live run of the same parser (language es; ingredients shortened to two):

{
  "recordType": "DISH",
  "name": "Desayuno estilo Cracovia",
  "originalName": "KRAKOWSKIE / CRACOW - STYLE BREAKFAST",
  "category": "Desayunos",
  "price": "34 zł",
  "composition": "Obwarzanek, pasta de huevo con cebollino, requesón con atún, hummus",
  "description": "Un desayuno variado que incluye un pan tradicional polaco, pasta de huevo, requesón con atún y hummus cremoso.",
  "ingredients": [
    {
      "name": "wheat flour",
      "display_name": "Harina de trigo",
      "explanation": null,
      "allergens": [
        {
          "name": "gluten",
          "confidence": "high"
        }
      ]
    },
    {
      "name": "egg",
      "display_name": "Huevo",
      "explanation": null,
      "allergens": [
        {
          "name": "eggs",
          "confidence": "high"
        }
      ]
    }
  ],
  "dietaryType": "Pescetarian",
  "dietaryTypeExplanation": "Este plato contiene atún, lo que lo clasifica como pescetariano. También incluye lácteos y huevos.",
  "allergens": "Gluten, Huevos, Leche, Pescado, Sésamo",
  "type": "dish",
  "isAlcoholic": false,
  "story": "El obwarzanek es un pan tradicional polaco, similar a un pretzel, originario de Cracovia y reconocido como producto regional protegido.",
  "nutritionInfo": {
    "serving_size_grams": "300g",
    "calories": "450-550 kcal",
    "protein": "25-35g",
    "fat": "15-25g",
    "carbs": "40-50g"
  },
  "sourceImageUrls": [
    "https://raw.githubusercontent.com/Exdenta/traveleat-support/main/docs/assets/menu-demo-small.jpg"
  ],
  "sourceImageRefs": [
    "https://raw.githubusercontent.com/Exdenta/traveleat-support/main/docs/assets/menu-demo-small.jpg"
  ]
}

Menu Parser API

Results are stored in a dataset that you can export as CSV, JSON or Excel, or read on demand through the Apify API.

Support and Feedback

We improve these Actors from your feedback. If something breaks or you need a field:

• Open an issue on the Actor's Issues tab
• If this Actor saved you time, a short review on the Store helps other users find it.

Run evidence

Start with one input and output

An Apify account is required. Use your Apify API token when calling the API; Console runs do not need that token pasted into the input. The unchanged default preview requires no provider key and makes no live AI or source call. A customized live run requires your own Gemini key; Maps/Outscraper collection also requires your own Outscraper key. Those providers bill separately.

Free-tier event rates checked 2026-10-04: actor-start $0.00005; result $0.0001.

Cost: Total = Actor event charges + platform usage if separately billed + any BYO-provider charges. See “Current costs” for your event rates, plan tiers and a calculation. A sample does not promise future billable volume.

This matching input/output pair is a marked DEMO preview, observed in run feRUIWM93MoYaIw7n (immutable build KGZ42ePpq0LTk11bR). It demonstrates the output shape and makes no live source or AI-provider call. It does not verify a live parse or synchronization. Keep the preview unchanged; customize it and supply the documented credentials for live work.

Input JSON

{
  "imageUrls": [
    "https://raw.githubusercontent.com/Exdenta/traveleat-support/main/docs/assets/menu-demo-small.jpg"
  ],
  "geminiModel": "gemini-2.5-flash",
  "language": "en"
}

Matching output excerpt

Long strings and arrays are shortened for display; this is not a complete dataset fixture.

{
  "recordType": "DEMO",
  "isDemo": true,
  "demoSource": "owned-default-prefill",
  "demoNote": "Deterministic preview only: no live source, photo scrape, or Gemini call occurred, and the row was not billed. Customize the input for live output; BYOK listings also require their provider credentials.",
  "name": "DEMO SAMPLE — Margherita Pizza",
  "originalName": "Margherita Pizza",
  "category": "Pizza",
  "price": "$18.00",
  "composition": "Tomato, mozzarella and basil",
  "description": "Example structured menu record; not a live extraction.",
  "ingredients": [],
  "dietaryType": "Vegetarian",
  "dietaryTypeExplanation": "Example classification for preview only.",
  "allergens": "Milk (preview example)",
  "type": "dish",
  "isAlcoholic": false,
  "story": null,
  "nutritionInfo": null,
  "sourceImageUrls": [
    "https://raw.githubusercontent.com/Exdenta/traveleat-support/main/docs/assets/menu-demo-small.jpg"
  ],
  "sourceImageRefs": [
    "https://raw.githubusercontent.com/Exdenta/traveleat-support/main/docs/assets/menu-demo-small.jpg"
  ]
}

Upload menu photos (or send URLs/base64 through JSON), get back the menu as structured JSON — one record per dish. Richer than generic menu-OCR actors:

• name (translated to your language) + originalName (as printed)
• category, price (with currency, as shown)
• ingredients — atomic base ingredients (composite preparations like pesto or hummus are decomposed), each with an English database name, translated display name, and EU-14 allergens with confidence levels
• dietaryType — Vegan / Vegetarian / Pescetarian / Other, with explanation
• nutritionInfo — estimated serving size, calories, protein, fat, carbs
• story — origin story / cultural significance
• isAlcoholic for beverages

15 output languages. The prompt is the production pipeline of a dietary-restrictions travel app — battle-tested on real menus worldwide.

Bring your own Gemini key

Parsing runs on your Google AI Studio key (get one free) — you control the AI cost and quota.

Check this Actor’s Pricing tab for the current charges for your Apify plan. Actor charges are separate from usage on your Gemini account. Your first live run waives up to 10 per-dish charges; the actor-start charge still applies. Empty and failed parses produce a diagnostic: true row with a message and create no Apify billing events, although Gemini may charge your account for an attempted inference.

Input example

{
  "imageUrls": ["https://example.com/menu-page-1.jpg", "https://example.com/menu-page-2.jpg"],
  "geminiApiKey": "AIza…",
  "language": "en"
}

Notes

• The untouched keyless Console run returns one deterministic row labelled recordType: "DEMO" / isDemo: true. It makes no image or Gemini call and is not billed. Customize the input and add your Gemini key to use the normal live path.
• The Console accepts local JPEG, PNG, or WebP uploads directly. API callers may use HTTP(S) URLs, base64 data URIs, or raw base64 in imageUrls.
• Each decoded image is limited to 10 MiB. Inline payloads and signed URL credentials are replaced with safe references in output records and logs.
• First live run allowance: up to 10 per-dish charges are waived; the actor-start charge still applies. The untouched prefilled DEMO is separate and completely free.
• Up to 20 images per run, parsed together as one menu.
• Unreadable items are skipped rather than guessed; empty results produce a non-billed diagnostic row explaining why.
• Don't have menu photos yet? The companion Google Maps Menu Scraper actor finds them from a Google Maps place URL and parses them in one run.

Safety

Allergen, dietary, ingredient and nutrition fields are AI-generated estimates. They can be wrong and do not account reliably for substitutions or kitchen cross-contact. Verify critical information with the restaurant; do not rely on this output as medical advice or as the sole safeguard for a severe allergy.

Support & links

• Questions, bug reports and FAQ: github.com/Exdenta/traveleat-support
• This actor runs the same menu pipeline as Travel Eat — scan a menu, get allergens, ingredients and dietary filters in 36 languages.

Product guides and related tools

• Product overview and public links
• Public setup guide
• Travel Eat menu API guide
• Menu API documentation and support
• Restaurant Dish Photo Matcher — Photos to Menu Items

Current costs

Current Actor event rates, checked 2026-10-04. USD; use the Pricing tab for your plan and later changes.

Event	Free	Bronze	Silver	Gold	Platinum	Diamond
Actor start (actor-start)	$0.00005	$0.00005	$0.00005	$0.00005	$0.00005	$0.00005
Dish extracted (result)	$0.0001	$0.0001	$0.0001	$0.0001	$0.0001	$0.0001

For example, 1 × actor-start, 100 × result totals $0.01005 in Actor event charges on Free. This is an arithmetic example, before any waiver; actual charged events depend on the run and selected options. It is not a promise of output or a charge for simply setting an item cap.

Your total is the charged Actor events, plus platform usage only if this Actor’s Pricing tab bills it separately, plus any external provider you use with your own key. Do not add platform usage twice when it is included. Apify credits and plan discounts affect your invoice. Set a run cost limit in Console before starting.

As documented, the first successful live run waives up to ten per-dish events; the actor-start event still applies. The untouched prefilled sample is a marked demo, not a live parse. Empty/failed diagnostic output creates no Actor billing events.

Agent setup skill

Use the maintained SKILL.md to configure the documented inputs, select latest, record the immutable build and distinguish source results from preview or diagnostic rows. Check the current Pricing tab before starting a run.

---

Independent tool. We built and run this Actor ourselves. Google does not operate, sponsor or endorse it, and its name is used only to say which service the data is processed with. We are not affiliated with Google.