GIF
GIF
the lovely us

🚗 RMTC: The Ultimate Digital Salesman

"We don't just rent cars. We rent feelings."

RMTC App Demo

💡 Inspiration

We realized that the rental car industry has a 80/20 problem. The 20% (Porsche collectors) know what they want. But the 80%, the masses, need guidance. They need a hero.

Currently, rental apps are just digital forms. They lack the charisma, the persuasion, and the psychology of a top-tier salesperson on the showroom floor. We wanted to clone the best car rental agent in the world, give them infinite scalability, and put them in your pocket.

🤖 What It Does

RMTC is a personalized, voice-and-vision AI agent that transforms the standard car pickup into a dynamic, high-touch video consultation.

1. The Hook (Deep Data Mining & Personalization) When a user opens the app, we go beyond simple dates. We analyze their rental history, physical appearance, and demographics to construct a psychological profile.

Visual Matching: The AI avatar dynamically shifts its appearance (gender, style, demeanor) to mirror the user’s demographic for maximum immediate rapport.
Active Memory: It interacts naturally by recalling past context (e.g., "How was your last trip to Mallorca?") and detecting real-time visual cues from the live camera feed.

2. The Consultation (The "Local Guide") The AI builds trust by acting as a concierge. If a user mentions they are visiting for specific activities (like nightlife or business), the AI recommends local hotspots and displays relevant visuals. It establishes value as a helpful consultant rather than just a rental clerk.

3. The "Tinder for Cars" (Visual Persuasion) We ditch static JPEGs. The AI presents vehicle upgrades using an addictive, interactive swipe interface.

The Anchor: It utilizes psychological anchoring by presenting standard options before revealing a premium "hero car" (e.g., a high-end sports car).
Generative Emotion: We use Veo 3 Fast to generate real-time videos of that specific car driving in the user's current destination. We sell the feeling of the drive, not the technical specs.

Tinder for Cars

4. The Upsell (Context-Aware Logic) Once the user is excited about the vehicle, we use real-time data to drive ancillary revenue like insurance.

The Logic: Our Algorithm checks the live weather forecast for the duration of the rental.
The Fear: If bad weather is predicted, we generate a video simulation of specific weather damage (e.g., a storm cracking a windshield) to visualize the risks.
The Close: The user visualizes the "why," significantly increasing the conversion rate for the protection package.

Weather-based Insurance Upsell

5. The Reward
The AI finalizes the booking and generates a QR code that the customer can scan directly at the Sixt machines to instantly pick up their keys.

Booking Finalization

6. Marketing Strategy
We leverage the existing Sixt–Payback partnership to offer customers a seamless 1% Payback bonus, creating a simple win-win for both sides. Payback

⚙️ How We Built It

We created a multimodal architecture to ensure the AI feels human and responsive.

Frontend: React with heavy WebRTC integration for real-time communication.
The Brain: OpenAI Realtime API handles the audio streams and provides computer vision (to see the user's reaction and accessories).
The Visuals: Veo 3 Fast generates context-aware car commercials on the fly (e.g., "Show me a Porsche outside a club in Ibiza").
Backend: FastAPI orchestration layer connected to a PostgreSQL database for user profiling.
Integrations: Google Maps for location consulting and the Sixt API (mocked) for inventory, WeatherApi, Hugging face.

System Architecture

🧠 Challenges We Ran Into

Latency: Making the AI interruptible and convincing over WebRTC was tough. We had to fine-tune the turn-taking logic so it didn't feel robotic.
The "Sales Persona": Prompt engineering the model to be "persuasive and confident" without being annoying required deep testing of the system prompt.
Video Gen Speed: Generating "dreams" (car videos) in real-time is computationally heavy; we optimized by pre-fetching common scenarios based on the user's destination.
DevPost banned us for some reasons XD
The slow Eduroam wifi, that didn't help at all

🏆 Accomplishments That We're Proud Of

Team Synergy: Met 2 amazing nerds, we harmonized, vibed and built together, despite not knowing each other beforehand.
Digital Charisma: The AI actually remembers small details (like the sunglasses). It feels like it cares.
The Visual Experience: Moving away from boring car lists to a "Tinder-style" video feed completely changes the user psychology.
Dynamic Upselling: Successfully using live weather data to sell insurance in a way that feels helpful, not predatory.

🚀 What's Next for SIXT x MRTC

AR Inspection: Allowing the user to walk around the car with their camera while the AI points out features.
Post-Rental Companion: The AI remains active during the trip as a concierge service (booking restaurants, finding parking).
US Expansion: Scaling the personality models to fit different cultural sales tactics.