MenuGen

About

MenuGen solves a small, very relatable problem: you sit down at a restaurant, open the menu and have no idea what half the dishes are. You take a photo of the menu, and the app reads every item and generates an illustrative picture of each one, laid out as a visual grid with the dish name and its description. It is for curious eaters, travellers facing menus in unfamiliar cuisines, and anyone who orders better when they can see the food. The app was made by Andrej Karpathy, the researcher who coined the term "vibe coding", and it is probably the most famous example of the practice. In his April 2025 essay "Vibe coding MenuGen" he explains that Cursor with Claude wrote 100% of the code, while he steered. The write-up is just as candid about the hard parts: the local prototype came quickly, but authentication, payments, API keys and deployment took most of the effort. That honesty made MenuGen a reference point for everyone shipping their first AI-built app, and the site still credits itself as "vibe coded at a hackathon by @karpathy". Highlights: • Upload or photograph any printed menu and get a picture for each dish • Menu text is read automatically with OCR, so there is nothing to type • Results are shown as a clean card grid with section, dish name and ingredients • Sign in with your Google account • Credits are bought through Stripe checkout, so there is no subscription to manage • Runs in any mobile or desktop browser, handy right at the table The images are illustrations generated by AI, not photos of what that particular restaurant serves, so treat them as a guide rather than a promise. Using MenuGen requires buying credits; Karpathy has said he keeps a 10% markup on top of the AI costs, but the site does not publish a public price list. It is a web app and works in any modern browser on phones and computers.

Reviews