Why I Built CookThis: An AI Recipe Generator From a Food Photo

You see a plate of food at a restaurant, in a TikTok video, or at a friend’s table and immediately think: “I want to make that.”

Then the friction begins.

You try to name the dish, open several recipe sites, dismiss pop-ups, compare conflicting ingredient lists, and decide whether any version will work with your cooking ability or the equipment in your kitchen. Sometimes you do all of that and still order takeout.

I built CookThis to make that moment simpler.

CookThis is a camera-first AI recipe generator from a food photo. Take a picture or choose one from the gallery, add any context you know, and CookThis creates an estimated home-cookable recipe with practical ingredients, clear steps, substitutions, and cooking guidance.

The idea is straightforward: see food, snap it, and cook your version.

The Problem Was Not a Lack of Recipes

The internet already has more recipes than anyone could cook in a lifetime. The problem is the gap between inspiration and action.

People photograph restaurant dishes, save food videos, bookmark recipe pages, and collect screenshots. Much of that inspiration is never cooked because turning a visual craving into a usable plan still requires work.

A search might return an authentic version that assumes specialist ingredients. A beginner recipe might still use twelve components and unexplained techniques. A generic AI answer might ignore the equipment, dietary rules, or regional availability that determines whether a recipe is useful.

CookThis focuses on that gap. It does not try to become a giant public recipe database, calorie tracker, grocery service, or full meal planner. Its job is narrower: turn food inspiration into a realistic starting point for a home kitchen.

How CookThis Turns a Photo Into a Recipe

Using CookThis starts with the food instead of a search box.

1. Take or choose a photo

Photograph a dish with the in-app camera or select an image, screenshot, or saved food photo from the gallery. Clear lighting and a useful angle give the system more visual context.

2. Add what the image cannot show

A photograph cannot reveal every seasoning or technique. Optional context can include the dish name, where it was found, its texture or flavor, servings, available equipment, and the kind of result you want.

Even a short note such as “spicy, creamy, served at a cafe in Karachi” can make an image-based estimate more useful.

3. Analyze in stages

CookThis does not rely on one enormous prompt. The pipeline separates food analysis, likely ingredients, recipe construction, and validation into focused stages.

That structure makes uncertain decisions easier to inspect and helps catch conflicts before a recipe reaches the user. It also makes the system easier to improve than a single opaque request that attempts to do everything at once.

4. Choose the version that fits

The same dish can lead to different recipes depending on the goal:

• Quick favors a shorter method and fewer moving parts.
• Easy explains more and uses beginner-friendly cues.
• Closest Taste spends more effort approximating the photographed dish.
• Healthier suggests lighter alternatives where they still make culinary sense.

There is no universally perfect recipe for a photo. The useful recipe is the one the person can actually make.

5. Move into cooking mode

Once the recipe is ready, cooking mode presents the method as a sequence rather than another long page to scroll through with messy hands. Recipes can be saved and reopened when the dish is worth making again.

Why Finished Food Is a Different Starting Point

Many photo-based recipe products scan a fridge or pantry and answer “What can I make with this?” That is useful, but it is a different job.

CookThis is designed first for “How can I make something like that?”

The user already has intent because the desired result is visible. They may not know the dish name, the cuisine, or the correct prompt, but they know what caught their attention. The camera preserves that intent without forcing the user to describe it perfectly.

CookThis also supports ingredients and cravings, but the finished dish remains the product’s clearest starting point.

A Home Version, Not a Secret Recipe

Restaurant kitchens may use commercial ovens, high-output burners, prepared sauces, specialist tools, or components made over several days. A social creator may omit an ingredient from the video. A photograph cannot expose what is hidden inside a sauce.

CookThis therefore does not claim to discover an exact restaurant or creator recipe. It creates a simple home version inspired by the available evidence.

That wording is deliberate. A useful approximation is more honest than an impossible promise of perfect reconstruction.

Recipes Adapted to the Cook

A beginner with one pan should not receive the same instructions as an experienced cook with an oven, air fryer, and pressure cooker.

CookThis can use preferences such as:

• cooking ability
• available equipment
• number of servings
• country or region
• recipe mode
• dietary preferences
• allergies and avoided ingredients

Region matters more than it first appears. Ingredient names, package sizes, and availability differ between Pakistan, the United Kingdom, and the United States. A recipe becomes more useful when it suggests an alternative a person can realistically buy.

Halal-Aware Without Pretending to Certify Food

Many recipe tools reduce halal preferences to “remove pork.” Real ingredient review can also involve bacon, ham, lard, gelatin, alcohol, cooking wine, and wine-based sauces.

CookThis can check generated recipes for common conflicts and suggest alternatives. It cannot certify a packaged product, supplier, or piece of meat. The user must still inspect labels and make the final dietary decision.

This boundary is important. Preference-aware generation should help people notice conflicts, not replace their judgment.

What About Inaccurate or Unsafe AI Recipes?

AI-generated recipes can be wrong. They may contain an implausible quantity, miss an ingredient mentioned in a step, suggest a poor substitution, or underestimate cooking time.

CookThis reduces those risks through separated generation stages, structured recipe data, dietary checks, ingredient conflict detection, food-safety reminders, and visible uncertainty. None of those measures turns AI into an unquestionable authority.

Users should review ingredients, verify allergies, check dietary suitability, and follow trusted guidance for safe internal cooking temperatures. The product should reduce work between inspiration and cooking, not ask anyone to switch off their judgment.

Can CookThis Use TikTok Food Inspiration?

On Android, CookThis supports certain shared TikTok food links when suitable preview information is available.

The title, creator information, and preview image are treated as clues. CookThis does not claim to download every video frame or recover a creator’s private method. If a link does not expose a useful preview, a gallery screenshot is usually the more reliable starting point.

The output remains an estimated home version inspired by the available context.

The Technology Behind CookThis

CookThis is built with Expo and React Native. React Native Vision Camera powers capture, while Convex manages backend data, recipe records, image storage, and live generation status. OpenRouter routes the staged AI work, Clerk handles identity, RevenueCat manages entitlements, and PostHog supports product analytics and AI observability.

The mobile code is organized by product feature so scanning, recipes, onboarding, billing, settings, and authentication can evolve without turning route files into application logic.

That organization became important as the product grew from a camera prototype into a system with uploads, AI generation, saved recipes, cooking mode, preferences, shared-link support, and subscriptions.

Who CookThis Is For

CookThis is for people who choose food visually. It is especially relevant if you:

• photograph restaurant or cafe dishes
• save food videos but rarely revisit them
• are learning to cook and find recipes intimidating
• want a quicker or simpler version of a dish
• need instructions adapted to limited equipment
• follow halal, vegetarian, vegan, or other dietary preferences
• want realistic substitutions for ingredients you cannot find
• prefer direct cooking guidance over recipe-site clutter

It is not a replacement for professional allergy advice, verified food-safety information, or a restaurant’s original recipe.

It is a bridge between “that looks good” and “I can try making that.”

Get CookThis on Google Play, or explore the full CookThis project for the product and engineering decisions behind the app.

frequently asked questions