maptourai

An AI-powered tour guide that combines real-world exploration with an engaging, educational experiences using AI avatars.

Comment 1

Inspiration

Vayable (YC S12) inspired us as a marketplace for local tour guides, offering more authentic travel experiences. We decided to reimagine this in 2025 with AI at its core, thus eliminating the marketplace friction that limited Vayable's scalability. While Vayable required advance booking, limited guide availability, and operated only in select cities, we built an AI-native tour guide that's available 24/7 for any location worldwide. Our vision is to transform every spontaneous walk into an immersive guided tour, making cultural education accessible to anyone, anywhere, instantly.

What it does

Maptourai is an intelligent, personalized tour guide that brings history and culture to life as you explore:

Real-Time Proximity Narration: When you're within 50 meters of a historical landmark, the app automatically detects your location and triggers AI-generated narrations which means no searching or booking required​

AI-Powered Location Discovery: Discovers nearby museums, monuments, historic sites, cathedrals, and cultural venues within 1.5km as you walk, ensuring you never miss a hidden gem​

Immersive Storytelling: Generates engaging 60-90 second video narrations with historical context, fascinating details, and hidden stories that make every location come alive. The video only unlocks when you're close enough, with the effect of encouraging you to explore and walk to each landmark​​

Interactive Map Experience: Beautiful map interface with tour guide markers, real-time user location tracking, multiple map styling options (streets, satellite, outdoors), and direct walking directions to guide you between locations​

Smart Caching: Remembers locations you've visited and stores narrations locally, so you can revisit stories instantly without waiting for regeneration​

Zero Planning Required: Unlike traditional tour apps that require advance planning or booking, Maptourai works spontaneously, you just open the app and start walking. The proximity-trigger on the AI avatar videos gamify exploration, motivating you to discover more locations as you go​

How we built it

Next.js App with TypeScript as Full Stack FastAPI async web framework for most Backend since Dedalus Labs only allows Python Mapbox GL JS for performant, interactive map shadcn/ui + Tailwind CSS for modern, responsive UI Dedalus Labs for multi-agent and MCP orchestration​ Agent 1 (Location Discovery): GPT-4.1 queries Foursquare for diverse location types (historic sites, museums, monuments, cathedrals, theaters, government buildings)​ Agent 2 (Narration Generation): Claude Sonnet 4 conducts web research via Brave Search API, synthesizing historical facts, architectural details, and cultural significance into 250-300 word scripts D-ID and ElevenLabs to create AI Avatar Tour Guides from Narration Scripts Saved video files on AWS S3 Buckets and links to them in SQLite

Challenges we ran into

iOS Device Orientation Permissions: iOS 13+ requires device orientation permission to be requested from a user gesture, not on mount. We refactored to trigger permission requests from button clicks​

API Rate Limiting Across Multiple MCPs: Foursquare and Brave Search have rate limits. Implemented batch processing with 2-second delays between narration generation requests and error handling with retry logic​

Incomplete AI Responses: Some Claude responses included research notes/preambles instead of clean narration scripts. Built narration cleaning functions with regex parsing to extract only the final script​

Location Diversity Problem: Initial searches returned too many similar locations (e.g., 4 fountains). Enhanced agent prompts to explicitly request diverse categories (museums, monuments, religious sites, etc.)​

Accomplishments that we're proud of

Multi-Agent Architecture: Successfully orchestrated two specialized AI agents (GPT-4.1 for data extraction, Claude Sonnet 4 for creative storytelling) working in a seamless pipeline​

Zero-Friction User Experience: The narration panel appears automatically when approaching markers, creating a "storytelling follows you" experience that is immersive with no buttons, no searching, just the opportunity to walk and learn​

Built an AI-native version of Vayable (YC S12), eliminating marketplace friction while preserving authentic local storytelling​

Generated narrations aren't just factual, they're story-driven with engaging hooks, hidden details, and contemporary relevance that make history come alive​

What we learned

Different AI models excel at different tasks. GPT-4.1's structured reasoning is perfect for data extraction, while Claude Sonnet 4's creative storytelling makes narrations engaging​ and they work together far more effectively than a single generalist agent

Technical capabilities mean nothing if the UX isn't delightful. Smooth animations, instant caching, and thoughtful loading states make AI feel seamless​

Guiding AI to produce consistent, high-quality narrations required iterative prompt refinement and specific structural guidance

The proximity-based unlock mechanic turned passive sightseeing into active exploration where users naturally want to "collect" location stories by walking to them​

What's next for maptourai

Launch on Product Hunt to attract early adopters, gather user feedback, and build a community of travel enthusiasts who can help shape the product roadmap​

Generate narrations in multiple languages using Eleven Labs and choose language based on user location

Save favorite locations, create custom tour routes, and share tours with friends which turns solo exploration into shared experiences​

Voice Commands with Tour Guides using D-ID for hands-free exploration while walking​

Make Free tier with 5 narrations/day; Premium $4.99/month for unlimited narrations, offline mode, and exclusive content

Built With

Share this project:

Updates