RecipAI — Snap a Photo, Get the Recipe

You've eaten it. You've seen it. You just don't know how to make it.

We built RecipAI for the moment that happens to everyone: a dish shows up in front of you — at a restaurant, on a friend's table, in a photo someone sent you — and you have no idea what's in it or how it was made. RecipAI closes that gap. Drop a photo, or take one on the spot, and it hands back the recipe.

Try RecipAI — recipai.cortexresearch.group

The premise

Recipes online are usually written by people who already know what they're making. RecipAI inverts that. It starts from the finished dish — the thing you can actually see — and works backward to the ingredients, quantities, and technique that most plausibly produced it.

No searching for a name you don't know. No scrolling past a life story to find the ingredient list. You upload a picture, and the picture is the query.

How it works

1. Drop a photo or take one live. Upload an existing image or use your camera right in the browser — no app install required.
2. AI reads the dish. A vision model identifies the dish, its likely cuisine, and the visible components — proteins, vegetables, sauces, garnishes, cooking method (fried, grilled, braised, raw).
3. The recipe gets reconstructed. From that visual breakdown, RecipAI generates a full ingredient list with estimated quantities, plus step-by-step instructions written the way a normal recipe reads — prep, cook, plate.
4. You cook it. Serving size and a few substitution notes are included, since a photo can't always tell you if something was gluten-free or dairy-free — RecipAI flags where it's inferring rather than certain.

Why this is harder than it looks

A photo doesn't show you salt, acid, or the three minutes something rested before it hit the pan. RecipAI is explicit about that instead of pretending otherwise: it distinguishes what it can see (a seared crust, a glossy reduction, char marks) from what it's inferring (likely a butter-finished pan sauce, likely rested 5 minutes). That distinction is what keeps the output usable instead of just confident-sounding.

The other hard part is genericity. A photo of fried rice could be a dozen different regional variations. Rather than guessing one and hiding the ambiguity, RecipAI names the most likely version and notes the near-neighbors it considered, so you can nudge it if it guessed wrong.

Who it's for

• Home cooks who see a dish somewhere and want to remake it that night, not after twenty minutes of searching for the right search term.
• Anyone building a recipe collection from restaurant meals and travel photos.
• People cooking around dietary constraints, who want a starting recipe they can then adapt.