Snap a photo of any restaurant menu and MenuGen uses AI to generate an image of every dish, so you can actually see what you are about to order.
MenuGen turns a photo of a restaurant menu into a visual menu: upload the menu image, and the app generates an AI image for each dish so you can decide what to order based on what the food actually looks like — not just an unfamiliar name in a foreign language. The tagline on the homepage is "Turn Menus into Magic."
The app is famously the side-project of Andrej Karpathy (@karpathy), co-founder of OpenAI, founding member of the Tesla Autopilot team, and former Director of AI at Tesla. He built MenuGen at a hackathon and wrote a now widely-cited essay about the experience on his blog ("Vibe coding MenuGen", karpathy.bearblog.dev). In the post and the launch tweet (x.com/karpathy/status/1917961248031080455), Karpathy describes using Cursor as the primary editor and explicitly calls out Claude as the model doing the heavy lifting on the codebase — making MenuGen one of the highest-profile public "vibe coded" apps from a top AI researcher.
Technically, MenuGen is a Next.js app on Vercel. Authentication is handled by Clerk (the live publishable key is wired to clerk.menugen.app), and the visual style is Tailwind-driven with purple-pink-orange gradients. Each dish image is generated on demand from the parsed menu text.
Beyond the product itself, MenuGen is interesting as a culture artifact: it helped popularize the term "vibe coding" and is one of the cleanest demonstrations that even very experienced engineers reach for AI editors and agentic coding tools to ship small consumer apps fast.
This listing is unclaimed. Sign in to claim it.