MenuMind

AI-powered menu visualization & translation — photograph any restaurant menu and instantly see a photorealistic image of every dish plus a translation in your language.

Get the app: Google Play · App Store

Live demo: https://menu-mind-tawny.vercel.app

Traveling abroad, you open a menu and recognize nothing. MenuMind lets you photograph any restaurant menu and instantly see every dish — translated into your language and rendered as a photorealistic AI image — so you actually know what you're about to order.

Demo

menumind-promo-45-traveler-tight.mp4

▶️ Watch the demo — photograph a menu and watch every dish get translated and rendered as an AI image.

Features

• 📸 Snap & translate — photograph any menu; every dish is extracted and translated (original name + your language), with the original-language text preserved.
• 🌍 40+ languages — handles even mixed-language menus (e.g. German + Italian on one page) without duplicate translations.
• 🎨 AI dish photos — a photorealistic image is generated for every dish (Flux diffusion model), so you can see an unfamiliar dish before ordering.
• ⚡ Progressive loading — translated text appears instantly while the AI images stream in (shimmer → fade-in); re-scanning a menu is served from cache immediately.
• 🥗 Dietary filters & tags — filter the menu by Vegetarian, Vegan, Gluten-free, Spicy or Sweet, with allergen warnings (⚠ gluten, nuts, dairy, …) surfaced per dish.
• 🔢 Nutrition estimates — calories plus protein / carbs / fat for each dish.
• 🗂️ Categorized menu + dish detail — dishes grouped by section (e.g. Antipasti, Mains); tap any dish for its photo, description, dietary info, nutrition and fun facts.
• 🕘 History — past scans are saved locally; rename, reopen or delete them.
• 🔗 Share — share a translated menu via link or QR code.
• 📱 Cross-platform — responsive web app and native Android & iOS apps: drag-and-drop or webcam capture on web, camera/gallery on mobile.

Tech stack

Backend

• FastAPI (Python), async SQLAlchemy
• PostgreSQL, Alembic migrations
• Containerized with Docker, deployed on AWS ECS

AI

• Google Gemini 2.5 Flash — vision LLM for menu extraction and translation
• FAL.ai Flux Schnell — diffusion model for dish image generation

Frontend

• Next.js 15, React 19, TypeScript
• Tailwind CSS, shadcn/ui
• Deployed on Vercel

AWS infrastructure

• ECS — backend compute
• RDS PostgreSQL — database
• S3 — image storage
• ECR — container registry

Architecture

flowchart LR
    User[User] -->|uploads photo| Vercel[Next.js on Vercel]
    Vercel -->|REST API| ECS[FastAPI on AWS ECS]
    ECS -->|extract & translate| Gemini[Google Gemini Vision]
    ECS -->|generate images| FAL[FAL.ai Flux]
    ECS -->|persist menu| RDS[(AWS RDS PostgreSQL)]
    ECS -->|store images| S3[(AWS S3)]
    Vercel -->|fetch images| S3

Loading