Video to Recipe API: YouTube Recipe Extractor & Nutrition

What does Video to Recipe API do?

Video to Recipe API turns cooking videos into clean, structured recipes: a title, every ingredient with its quantity and unit, numbered steps, servings, prep and cook times, an estimated nutrition per serving and an aisle-grouped shopping list. Give it YouTube links, direct audio or video file links, or captions you pasted from TikTok or Instagram Reels, and get one JSON recipe per video in your language and unit system.

Use it in Apify Console with no code, or as a recipe extractor API for your meal planner, grocery app, food blog or AI assistant.

Why use Video to Recipe API?

Recipes in short videos are hard to reuse: the amounts are spoken, half of them are in the description, servings are rarely stated and nothing is in grams. Video to Recipe API does the boring part for you:

• 🧾 Structured ingredients: quantity, unit, preparation note and the original line from the video, so you can check every value against the source.
• ⚖️ Servings scaling: set servings and every amount is scaled for you (2 → 4 servings doubles everything).
• 📏 Metric or imperial: amounts converted to grams and millilitres or to ounces, pounds and cups, rounded the way cooks measure (1/2 cup, 1 1/2 tbsp).
• 🥗 Nutrition per serving: calories, protein, fat and carbs estimated from USDA FoodData Central, with the matched USDA food for each ingredient.
• 🛒 Shopping list by aisle: duplicates merged (1 tbsp + 1 tsp soy sauce becomes one line), water left out, grouped in store-walk order with aisle names in your language.
• 🌏 12 output languages: English, Chinese (Simplified), Malay, Indonesian, Thai, Vietnamese, Japanese, Korean, Spanish, French, German and Portuguese, whatever language the video is in.
• 🔍 Honest about gaps: missingInfo lists what the video does not say (servings, times, amounts), and confidence tells you how complete the recipe is. Nothing is invented: no amount is guessed.
• 🤖 Apify platform features: run it by API, on a schedule, or send results to Google Sheets, Make, Zapier or a webhook with integrations.

Who uses it: meal planner and grocery app developers, food bloggers and recipe sites, nutrition and diet coaches, and anyone who saves recipes from YouTube Shorts, TikTok and Reels.

What data does Video to Recipe API return?

Field	Description
title, cuisine	Dish name in your output language, and the cuisine (for example malaysian)
servings, originalServings, scaleFactor	Servings of the result, of the source recipe, and the scaling applied
prepMinutes, cookMinutes, totalMinutes	Times stated in the video, or estimated from the steps (see estimatedFields)
ingredients	Name, English name, quantity + unit, normalizedQuantity + normalizedUnit + normalizedText, preparation, grocery aisle, optional flag, estimated grams, USDA match and the original line
steps	Numbered steps with a duration when the video states one
nutritionPerServing	kcal, protein_g, fat_g, carbs_g, how many ingredients were matched, and the data source
shoppingList	Aisles in store-walk order, each with merged items and display amounts
confidence, missingInfo, estimatedFields, warnings	How complete the recipe is and what is missing or estimated
sourceUrl, sourceType, dataSources	Where the recipe came from and which texts were read
youtube, media	Video ID and channel for YouTube links; duration and billed extra minutes for media files

How to extract a recipe from a video

1. Click Try for free and open the Input tab.
2. Paste YouTube video links, audio or video file links, or put pasted captions in Captions and mixed items (the example caption works as it is).
3. For YouTube links, add your own free YouTube Data API key (see the FAQ for the 2-minute setup).
4. Optionally choose Scale to servings, Units and Output language.
5. Click Start. A caption or YouTube recipe takes about 20–40 seconds; a media file takes longer because the speech is transcribed first. Open the Output tab for the recipes, nutrition and shopping lists.

How much does it cost to turn a video into a recipe?

Video to Recipe API uses pay-per-event pricing:

• $0.04 per recipe returned. This includes up to 3 minutes of audio or video.
• $0.015 per extra started minute of a media file beyond those 3 minutes.
• A start fee of $0.00005 per run. There is no charge for compute time.

Examples: 10 recipes from captions or YouTube links cost $0.40. A 10-minute video file costs $0.04 + 7 × $0.015 = $0.145.

• Failed items are free: a video that is not a recipe, a link that cannot be read or a file that is too long is not charged.
• Maximum recipes is a hard cap, and Apify stops the run at the maximum cost you set. The extra minutes of a media file are reserved from your remaining budget first; a file longer than the budget covers is skipped without charge.
• On the Apify free plan, a run returns up to 3 recipes and accepts media files up to 5 minutes.

Input

See the Input tab for every option. Example input:

{
    "youtubeUrls": ["https://www.youtube.com/watch?v=VIDEO_ID"],
    "youtubeApiKey": "YOUR_YOUTUBE_API_KEY",
    "mediaFileUrls": ["https://example.com/videos/pad-thai.mp4"],
    "items": [
        {
            "captionText": "Garlic butter shrimp (serves 2)\n300g shrimp, peeled\n3 tbsp butter\n4 cloves garlic, minced\n...",
            "sourceUrl": "https://www.tiktok.com/@example/video/123"
        }
    ],
    "servings": 4,
    "unitSystem": "metric",
    "outputLanguage": "en",
    "includeNutrition": true,
    "maxRecipes": 100,
    "maxMediaMinutes": 20
}

• items: one entry per recipe. An entry is an object with any of youtubeUrl, mediaFileUrl, captionText and sourceUrl, or a plain string (a YouTube link, a media file link or the caption text). When an entry has several sources, they are merged into one recipe (for example a YouTube description plus your pasted caption). Through MCP, use the object form ({"captionText": "..."}): MCP clients reject plain strings in this list.
• sourceUrl is kept in the output for your reference and is never opened.
• mediaFileUrls must be direct links to files you have the right to process (mp4, webm, mov, mp3, m4a, wav and other formats ffmpeg reads), up to 200 MB. Links to YouTube, TikTok, Instagram, Facebook and other platform pages or CDNs are refused.
• youtubeApiKey is stored as a secret input and is used only to call the official YouTube Data API.

Output

You can download the dataset in various formats such as JSON, HTML, CSV or Excel. A real result for the example caption above, scaled from 2 to 4 servings (ingredient, step and shopping-list arrays shortened here):

{
    "inputIndex": 0,
    "sourceUrl": "https://www.tiktok.com/@example/video/123",
    "sourceType": "caption",
    "title": "Garlic butter shrimp",
    "servings": 4,
    "originalServings": 2,
    "scaleFactor": 2,
    "prepMinutes": 5,
    "cookMinutes": 5,
    "totalMinutes": 10,
    "ingredients": [
        {
            "name": "shrimp",
            "nameEn": "shrimp",
            "quantity": 600,
            "unit": "g",
            "normalizedQuantity": 600,
            "normalizedUnit": "g",
            "normalizedText": "600 g",
            "preparation": "peeled",
            "groceryAisle": "seafood",
            "optional": false,
            "estimatedGrams": 600,
            "nutritionMatch": { "fdcId": 175179, "description": "Crustaceans, shrimp, raw", "dataType": "SR Legacy" },
            "originalText": "300g shrimp, peeled"
        },
        {
            "name": "butter",
            "quantity": 6,
            "unit": "tbsp",
            "normalizedText": "6 tbsp",
            "groceryAisle": "dairy_eggs",
            "estimatedGrams": 85,
            "originalText": "3 tbsp butter"
        }
    ],
    "steps": [
        {
            "order": 1,
            "text": "Melt 2 tbsp butter in a pan, cook the shrimp 1–2 minutes per side until just cooked through, then transfer the shrimp to a plate and set aside.",
            "durationMinutes": 4
        }
    ],
    "nutritionPerServing": {
        "kcal": 294,
        "protein_g": 30.9,
        "fat_g": 18.2,
        "carbs_g": 3.3,
        "estimated": true,
        "matchedIngredients": 5,
        "totalIngredients": 5,
        "source": "USDA FoodData Central (...), public domain (CC0)"
    },
    "shoppingList": [
        {
            "aisle": "produce",
            "aisleLabel": "Produce",
            "items": [
                { "name": "garlic", "quantity": 8, "unit": "clove", "amountText": "8 clove", "optional": false },
                { "name": "lemon", "quantity": 1, "unit": "piece", "amountText": "1", "optional": false }
            ]
        }
    ],
    "confidence": 0.96,
    "missingInfo": ["prepMinutes", "cookMinutes"],
    "estimatedFields": ["prepMinutes", "cookMinutes"],
    "warnings": [],
    "language": "en",
    "unitSystem": "metric",
    "media": null,
    "extractionVersion": 1
}

Each run also saves a SUMMARY record in its key-value store: items received, recipes delivered and failed, extra media minutes charged, the limit that applied, and the reason for each failed item.

Use Video to Recipe API as a recipe API

curl -X POST "https://api.apify.com/v2/acts/messnomore~video-to-recipe/run-sync-get-dataset-items" \
  -H "Authorization: Bearer YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"items": ["2 eggs, 2 tomatoes, 1 tsp sugar, salt. Fry the eggs, add the tomatoes and sugar, season."], "servings": 2, "outputLanguage": "en"}'

See the API tab for clients in JavaScript and Python. You can also run it from AI assistants through the Apify MCP server.

Tips for better results

• Give the YouTube link and paste the pinned comment or caption in captionText of the same item when the amounts are only in a comment: all texts are merged into one recipe.
• For TikTok and Reels, copy the caption text (and the video link as sourceUrl). If the amounts are only spoken, give a direct link to the video file in mediaFileUrl.
• Check missingInfo before cooking: quantity:<ingredient> means the video never said how much.
• Use servings to normalize every recipe to the same household size for a weekly meal plan.
• When a video does not say how many it serves, the servings are estimated from typical single portions (about 120–150 g meat, 80–100 g dry pasta or 2 eggs per person) and scaling starts from that estimate. The record says so in warnings and estimatedFields, so check it before you shop.

Related guides

• How do you turn a cooking video into a written recipe with amounts, steps and nutrition?
• 中文指南：做菜视频怎么转成带用量、步骤和营养的文字菜谱？

FAQ

How does the recipe extraction work?

The Actor reads the texts you give it: the YouTube title and description (through the official YouTube Data API), the speech of a media file (transcribed with OpenAI speech-to-text) and your pasted caption. An AI model (OpenAI) turns them into a structured recipe, copying each ingredient line from the source so you can check it. Amounts are then scaled and converted by code, and each ingredient is matched to a USDA FoodData Central food for the nutrition estimate (text embeddings find candidates, and a small AI model picks the closest one).

Which platforms are supported? Can I paste a TikTok or Instagram link?

• YouTube videos, Shorts and live links, with your own YouTube Data API key.
• TikTok, Instagram Reels, Facebook, Xiaohongshu and others: paste the caption text. The Actor does not open pages on those platforms or download from them.
• Any direct audio or video file you have the right to process, such as files from your own storage or CDN.

Why do I need my own YouTube API key?

The YouTube Data API terms require each developer to use their own key. A key is free: in Google Cloud Console, create a project, open APIs & Services → Library, enable YouTube Data API v3, then Credentials → Create credentials → API key. Each video uses 1 of your 10,000 free daily quota units. The Actor reads only the public details of the videos you list (title, description, channel name, publish date and duration); it does not download YouTube videos or captions.

How accurate is the nutrition estimate?

It is an estimate. Each ingredient is matched to a USDA FoodData Central food, its weight is computed from the amount (USDA portion weights for cups and pieces), and the totals are divided by the servings. Ingredients without an amount, marked optional, or used for deep-frying are left out, and nutritionPerServing is null when fewer than half of the ingredients have both an amount and a match. Weights of bone-in meat ("1.2 kg whole chicken") count the bones as meat, so those recipes show more calories than you eat.

Which languages are supported?

Videos and captions can be in any language the AI and the transcription understand, including English, Chinese, Malay, Indonesian, Thai, Vietnamese, Japanese and Korean. The output is written in one of the 12 output languages; Chinese output always uses Simplified characters. When a Japanese or Korean source leaves a few words untranslated, the record carries a warning.

Why did I get fewer recipes than videos?

Items that are not recipes (a vlog, an ad, "recipe in the comments"), links that cannot be read, and media files that are too long, too large or silent are skipped and not charged. The SUMMARY record lists each one with its reason. A run also stops at Maximum recipes, at your maximum cost per run, on the free plan after 3 recipes, and early when many items in a row fail (for example several long media files that contain no recipe).

Is my input stored or shared?

Your captions, the YouTube texts and the speech of your media files are sent to the OpenAI API only to transcribe them, extract the recipe and match the ingredients to USDA foods. Media files are downloaded to a temporary folder for the run and deleted right after. The Actor does not write your input or your API key to its logs, and it keeps no cache of YouTube data. Like every Apify run, your input and results stay in your own run storage, which you can delete at any time.

Is it legal to extract recipes from videos?

You decide what the Actor reads: only the YouTube videos you list (public title and description through the official API, under the YouTube Terms of Service), files you have the right to process, and text you paste. YouTube data is handled under the Google Privacy Policy; you can stop the Actor's access at any time by removing the key from your input or deleting it in Google Cloud Console. If you keep results from YouTube videos for more than 30 days, refresh or delete them as the YouTube API Services Terms require. A list of ingredients is generally not protected by copyright, but the creator's wording, photos and videos are: credit and link the original creator when you publish a recipe. Video to Recipe API is not affiliated with YouTube, Google, TikTok or Meta.

Privacy policy

The Actor processes the input of each run only to produce that run's results. Third-party processors: the OpenAI API (extraction and transcription) and, for YouTube links, the YouTube Data API with your own key. The developer does not receive, sell or share your input or results; they stay in your Apify account under Apify's privacy policy. Questions: open the Issues tab.

I found a wrong amount or a missing ingredient. What can I do?

Open the Issues tab and include the video link and the run ID. Feedback and feature requests are welcome there too.