Generation Runner is a conceptual virtual reality (VR) fitness application designed to transform indoor cardio (running, cycling, paddling) into an epic, time-traveling adventure. By blending generative AI environments with gamified, personalized workouts, athletes can experience exercising across different historical eras and spectacular locations.
The athlete’s real-world physical effort drives a 3rd-person avatar through a dynamically generated 3D world. As the user works out, they hit milestones, collect points, and pass through portals to teleport across time and space.
Key Gamification Mechanics:
- Dynamic HUD: Real-time tracking of distance, pace, heart rate, calories burned, and points.
- Level Progression: Reaching distance milestones triggers a stargate/portal, seamlessly leveling up the player and transitioning them to a completely new historical era.
- Inventory & Upgrades: Gathering points unlocks era-appropriate gear. For example, a runner might start barefoot, unlock prehistoric fur boots, grab a sweat-wicking towel in Egypt, and eventually earn modern branded racing shoes.
- Audio Feedback: A deep-voiced announcer calls out milestones (e.g., "ONE MILE!", "2 MILE DOWN!") and rewards to keep the athlete motivated.
The initial visual style and environmental concepts were established using Nano Banana 2. These early, somewhat less photorealistic screenshots successfully mapped out the foundational eras: Prehistoric jungles, Ancient Egypt, Roman Rome, Medieval London, the Wild West, and Victorian New York.
To prove the dynamic transition and progression, we utilized Google Flow alongside Google Veo 3 (Lite, Fast, and Quality models) to generate a series of continuous video scenes.
-
Iteration A: Detailed World-Building Focused on hyper-detailed photorealism and complex HUD retention. The generation successfully produced stunning early levels but encountered context limits and got stuck after the 4th scene (Rome).
📺 Watch Iteration A on YouTube -
Iteration B: The Seamless Journey (Submission) By streamlining the prompt wording, we successfully generated a continuous, multi-level sequence showcasing the core loop: running, collecting boosters, and teleporting through time. While the AI gradually lost the static HUD layout after the 3rd scene, we embraced this as a feature—a "Full Immersive Mode" where the athlete stops focusing on the numbers and fully loses themselves in the breathtaking VR experience.
📺 Watch the Full Journey on YouTube
Based on our generated scripts, a typical workout flows as follows:
- Lvl 1 - Prehistoric: Dodge dinosaurs, pass distant volcanoes, and earn your first "fur boots."
- Lvl 2 - Ancient Egypt: Run past the Pyramids and collect a sweat-wicking towel.
- Lvl 3 - Roman Empire: Sprint past the Colosseum and aqueducts.
- Lvl 4 - Medieval London: Navigate cobblestone streets with merchants to unlock a runner's shirt.
- Lvl 5 - Wild West: Dash through a frontier town to earn running pants.
- Lvl 6 - Modern / Future: Emerge in Shanghai or New York, finally equipping high-tech, modern branded racing shoes.
The next evolution of Generation Runner will leverage advanced world models, such as the Google Genie 3 project, to generate fully traversable 3D environments on the fly. This will allow the world to react seamlessly to user preferences, real-time pace, and biometric data.
The platform offers massive potential for organic, non-intrusive brand partnerships that actually enhance the world-building:
- Digital Apparel & Gear: Sports, hydration, and nutrition brands can offer their real-world shoes, kits, and accessories as unlockable in-game items.
- Immersive Race Promotions: A major event (e.g., the New York Marathon) could sponsor a generated NYC course, allowing users globally to run the route in VR.
- Branded Environments: A cycling company could sponsor a breathtaking Alps course. The scenery could be dotted with natural brand messaging—such as logos on race banners, flags, and billboards—making the athlete feel like they are in a real, sponsored race.
See all prompts, materials, and source code at: GenerationRunner on GitHub